Skip to content

per-artifact-methodologies.md

Latest commit

 

History

History
1094 lines (675 loc) · 86.1 KB

per-artifact-methodologies.md

File metadata and controls

1094 lines (675 loc) · 86.1 KB

Hack23 Logo

🧩 Per-Artifact Methodologies

📊 Construction Rules for Every Analysis Artifact Produced Under analysis/daily/*/
🎯 One Section per Artifact · Purpose · Inputs · Structure · Mermaid · Depth Floor

Owner Version Effective Date Classification

📋 Document Owner: CEO | 📄 Version: 1.4 | 📅 Last Updated: 2026-04-25 (UTC) 🔄 Review Cycle: Quarterly | ⏰ Next Review: 2026-07-31

🎯 How to Use This Document

Each section is self-contained and describes a single analysis artifact. Use it as a lookup:

  1. Artifact Catalog points you here for the construction rules of each artifact.
  2. Each section gives you: Purpose · EP MCP inputs · Required sections · Mandatory Mermaid · Depth floor · Quality signals.
  3. Apply the rules, write Pass 1, then read the whole file back and do Pass 2 (see ai-driven-analysis-guide.md §Step 9).

♻️ Re-Run Improve/Extend Semantics

When a same-day analysis folder already has a manifest.json.history[] entry from a prior run, re-running a workflow on the same date+type is never a no-op: every artifact must be extended or rewritten. The prior-run diff helper (scripts/aggregator/prior-run-diff.js, always-on) classifies each artifact as:

  • carryForward — artifact exists on disk, meets its line floor, has no placeholder markers, and (where required) contains a Mermaid block. Stage B MUST extend and deepen it — skip-writes are forbidden. The entry exposes priorLines (current on-disk size) and extendFloor (= max(threshold floor, priorLines + 20)). Stage B must log a line per artifact: [EXTEND-FROM-PRIOR: <relativePath> prior=<priorLines>L → new=<newLines>L (+<delta>)]. manifest.json.artifactSources["<relativePath>"] is set to "extend-from-prior:<priorRunId>".
  • rewrite — artifact is missing, below floor, contains placeholders, or lacks a required Mermaid block. Stage B rewrites it with a stronger version sized to the catalog floor.

Stage C always validates all artifacts, including those extended from prior runs. If an extended artifact does not reach its extendFloor (from prior-run-diff.json), Stage C flags it as a hard-RED extend:below-extendFloor violation. On re-runs, manifest.pass2.rewriteCount === 0 is also a hard-RED rerun-no-op violation. See scripts/validate-analysis-completeness.js for the enforcement logic.

Shared Hack23 Mermaid theme block — prepend this to every diagram:

%%{init: {"theme":"dark","themeVariables":{"primaryColor":"#1565C0","primaryTextColor":"#ffffff","primaryBorderColor":"#0A3F7F","lineColor":"#90CAF9","secondaryColor":"#2E7D32","secondaryTextColor":"#ffffff","tertiaryColor":"#FF9800","tertiaryTextColor":"#000000","mainBkg":"#1565C0","secondBkg":"#2E7D32","tertiaryBkg":"#FF9800","noteBkgColor":"#FFC107","noteTextColor":"#000000","errorBkgColor":"#D32F2F","fontFamily":"Inter, Helvetica, Arial, sans-serif"}}}%%

Color semantic mapping (use consistently):

Role in diagram Color
Primary actor / input / scope #1565C0 blue
Synthesis / cognitive layer #7B1FA2 purple
Safe / supporting / approved #2E7D32 green
Caution / medium risk #FF9800 orange
Threat / high risk / rejected #D32F2F red
Metadata / pending / note #FFC107 yellow
Reference / read-only #0288D1 light blue

🕵️ Cross-Cutting Tradecraft Reference

Every artifact in this document applies the professional standards defined in osint-tradecraft-standards.md:

  • Source grading — every citation carries an Admiralty grade (A1–F6); grade → 🟢/🟡/🔴 mapping per §2.3.
  • Estimative language — every probabilistic claim uses a Words-of-Estimative-Probability band (§3.1) with an explicit time horizon (§3.4). Banned ambiguous terms (§3.2) do not appear in analytic conclusions.
  • Structured Analytic Techniques — the run's methodology-reflection.md attests ≥10 SATs drawn from §4; each artifact below declares its Mandatory SATs. line (grep-able: ^\*\*Mandatory SATs\.\*\*) listing the required primary and supporting techniques. Techniques absent from an artifact's Mandatory SATs. line may still be applied at analyst discretion but are not checked at Stage C.
  • ICD 203 standards — the nine ODNI analytic standards (§1) map onto the sections of each artifact and are self-audited in methodology-reflection.md §12.
  • OSINT scope — sources outside §5.2 (personal-life data, doxing, covert collection, unverified leaks) are never cited.

Artifacts marked below with "Source grade × WEP discipline (tradecraft)" make probabilistic claims and therefore apply the full §2 + §3 discipline.

🧠 intelligence/

analysis-index

Purpose. Read-me-first index naming every artifact in this run and the recommended reading order for a downstream consumer (article generator, reviewer, next-run agent).

EP MCP inputs. None directly — built from manifest.json + filesystem walk of the run root.

Required sections.

  1. Run header — date, article type, run number, runtime, data sources attempted, data sources succeeded.
  2. Stage table — Stage 1 (Facts: classification/, documents/), Stage 2 (Quantification: risk-scoring/), Stage 3 (Intelligence: intelligence/), Stage 4 (Decision: synthesis-summary).
  3. One row per artifact with: path, purpose (one line), reading priority (P1/P2/P3), line count, status (complete / short / placeholder).
  4. Recommended reading order (numbered list) and expected total reading time.
  5. "Where to start if you only have N minutes" shortcuts for 5 / 15 / 30 minutes.

Mandatory Mermaid. flowchart LR with four stage nodes (blue → green → purple → orange) linking into the artifacts produced in each stage.

Depth floor (breaking): 160 lines.

Quality signals. Every path in the table resolves to an existing file; every file listed in manifest.files.* appears in the index; no "TBD" rows.

synthesis-summary

Purpose. The run's intelligence executive summary. The one file a reader consults if they only have 5 minutes.

EP MCP inputs. Consolidates every other intelligence/ and risk-scoring/ artifact; reads no raw MCP data directly.

Required sections.

  1. Executive finding — one paragraph naming the single most politically significant finding of the run with confidence level.
  2. Top-5 findings — table: finding, evidence (artifact citations), confidence (🟢/🟡/🔴), change vs. prior run.
  3. Parliament status dashboard — coalition stability, pipeline health, threat posture (each with a trend arrow).
  4. Stakeholder snapshot — 3–5 named actors with position shift vs. prior run.
  5. Forward monitors — ≥6 specific, time-bounded watchpoints (date, event, what-to-watch, trigger threshold).
  6. Confidence ledger — where confidence is HIGH vs. where LOW, and what evidence would change it.

Mandatory Mermaid. mindmap rooted at the headline finding with six branches (Coalition · Risk · Stakeholders · Economic · Threats · Forward), plus a small flowchart LR showing the forward-monitor timeline.

Depth floor (breaking): 205 lines.

Mandatory SATs. Key Assumptions Check (§4.1 technique 2, primary — surface structural assumptions about coalition arithmetic and EP calendar) · Quality of Information Check (§4.1 technique 3, primary — populate confidence ledger) · Scenario Analysis (§4.1 technique 9, supporting — inform forward-monitor watchpoints). See osint-tradecraft-standards.md §4.

Quality signals. Every claim in the Executive Finding paragraph links to at least one other artifact in the run; the Top-5 table has ≥3 numeric citations (score, margin, %); no paragraph exceeds 150 words without structure. Source-grade × Estimative-language discipline (tradecraft): every headline judgement uses a WEP band (see osint-tradecraft-standards.md §3.1) with explicit time horizon (§3.4); every Top-5 finding carries an Admiralty source grade (§2) and a 🟢/🟡/🔴 confidence marker consistent with §2.3.

stakeholder-map

Purpose. Named-stakeholder map placing ≥12 actors (political groups, key MEPs, committees, Commission DGs, Council configurations, external lobbies, citizen groups) on a Power × Alignment grid for the period's dominant issue.

EP MCP inputs. get_meps, get_committee_info, analyze_country_delegation, assess_mep_influence, get_speeches (for position attribution).

Required sections.

  1. Issue frame — one paragraph stating the exact question the map answers.
  2. Actor roster — table of ≥12 actors with: name, role, power score (0–10, justified), alignment (-5 … +5, justified).
  3. Power × Alignment quadrantChart (Mermaid) with every actor positioned.
  4. Quadrant narratives — one paragraph per quadrant (Champions, Defenders, Critics, Sceptics) with ≥150 words.
  5. Movement since prior period — arrows / table showing actor shifts.
  6. Coalition implications — which quadrant combinations can form majorities on this issue.

Mandatory Mermaid. quadrantChart with axes Power (y) and Alignment (x), ≥12 labelled points.

Depth floor (breaking): 305 lines.

Mandatory SATs. Stakeholder Mapping (§4.2 supporting, primary — power × alignment grid construction) · ACH (§4.1 technique 1, supporting — when ≥2 plausible alignment scores compete for the same actor). See osint-tradecraft-standards.md §4.

Quality signals. Every power score cites a source (MEP influence index, committee role, group seat share); every alignment score cites a roll-call, speech, or public position.

📎 Worked example: examples/stakeholder-map-good-output.md — illustrative 14-actor roster with cited power/alignment scores, four ≥150-word quadrant narratives, and a movement-since-prior-period table.

scenario-forecast

Purpose. Three to five forward scenarios with probability weights, early-warning indicators, and trigger events.

EP MCP inputs. track_legislation, monitor_legislative_pipeline, analyze_coalition_dynamics, get_plenary_sessions (upcoming), get_procedure_events.

Required sections.

  1. Horizon statement — timeframe covered (next 7 / 30 / 90 days).
  2. Baseline assumption — the single most important current-state claim the scenarios branch off.
  3. Scenarios — ≥3, each with: name, probability %, narrative (≥150 words), early-warning indicators (≥3), trigger events (≥2 date-bounded), stakeholder impact summary.
  4. Cross-scenario sensitivity — a single variable whose movement flips probability weights.
  5. Monitoring plan — what to re-check and when.

Mandatory Mermaid. flowchart TD branching from the baseline node into the named scenarios, then into early-warning indicators and trigger events. Use green for the most-likely branch, orange for medium, red for tail-risk branch.

Depth floor (breaking): 280 lines.

Mandatory SATs. Scenario Analysis (§4.1 technique 9, primary) · Pre-Mortem (§4.1 technique 8, primary — mandatory before finalising probability weights) · Key Assumptions Check (§4.1 technique 2, supporting — baseline assumption section) · Indicators & Signposts (§4.1 technique 4, supporting — early-warning indicators per scenario). See osint-tradecraft-standards.md §4.

Quality signals. Probabilities sum to 100%; each scenario names at least one procedure ID, date, or named actor; no scenario is described in bullet points alone. Source-grade × Estimative-language discipline (tradecraft): each scenario's probability is expressed as a WEP band (osint-tradecraft-standards.md §3.1) with explicit time horizon (§3.4); primary SAT is Scenario Analysis with a mandatory Pre-Mortem (§4 techniques 9 + 8); every indicator cites its source with an Admiralty grade (§2).

📎 Worked example: examples/scenario-forecast-good-output.md — three scenarios with WEP bands separated from confidence labels, dated indicators, a named Pre-Mortem scenario, and a cross-scenario sensitivity flip.

pestle-analysis

Purpose. Political / Economic / Social / Technological / Legal / Environmental scan of factors shaping the period's dominant issue.

EP MCP inputs. get_procedures, get_adopted_texts, search_documents, plus IMF economic context (sole authoritative source) and World Bank non-economic context (health, education, social, environment, demographics, defence, agriculture, innovation, governance only) via imf-indicator-mapping.md and worldbank-indicator-mapping.md.

Required sections.

  1. Issue frame — the question the scan answers.
  2. Six dimensions — one subsection each (Political, Economic, Social, Technological, Legal, Environmental), each with ≥150 words, ≥2 evidence citations, and a pressure rating 🟢/🟡/🔴.
  3. Pressure synthesis — which dimensions reinforce each other, which offset.
  4. Implications for EP — how the scan translates to committee / coalition / procedure action.

Mandatory Mermaid. mindmap with six coloured branches (Political=blue, Economic=orange, Social=green, Technological=purple, Legal=light-blue, Environmental=red) and named sub-drivers per branch.

Depth floor (breaking): 250 lines.

Mandatory SATs. PESTLE (§4.2 supporting, primary — the six-dimension scan is itself the SAT) · Force-Field Analysis (§4.2 supporting, supporting — pressure synthesis section). See osint-tradecraft-standards.md §4.

Quality signals. At least one IMF indicator cited for any economic dimension; WB non-economic indicator additionally cited for health/education/social/environment/defence dimensions; legal dimension cites at least one treaty article or CJEU reference; each dimension's pressure rating has a written justification.

threat-model

Purpose. Multi-framework threat view of the period — Diamond Model + Attack Trees + Kill Chain — applied to democratic / institutional threats (never software-centric).

EP MCP inputs. get_meps, get_parliamentary_questions (accountability probes), get_mep_declarations (interest exposure), get_voting_records.

Required sections.

  1. Threat landscape summary — 6 purpose-built dimensions per political-threat-framework.md.
  2. Diamond Model — per top threat: adversary, capability, infrastructure, victim (named).
  3. Attack tree — root goal → subgoals → specific actions (named procedures, named MEPs where relevant).
  4. Kill chain — reconnaissance → weaponisation → delivery → exploitation → objectives, mapped to political manoeuvres.
  5. Mitigation posture — existing institutional safeguards and their residual gaps.

Mandatory Mermaid. graph TD attack tree with top node red, intermediate nodes orange, leaf actions yellow. Arrows labelled with the action.

Depth floor (breaking): 250 lines.

Mandatory SATs. Key Assumptions Check (§4.1 technique 2, primary — threat assumptions) · Red Team / Devil's Advocate (§4.1 technique 7, primary — ≥1 alternative hypothesis per headline threat) · ACH (§4.1 technique 1, supporting — when ≥2 threat hypotheses compete). See osint-tradecraft-standards.md §4.

Quality signals. Each threat is named, not generic; mitigation posture references at least one EP rule, treaty article, or institutional practice. Source-grade × Estimative-language discipline (tradecraft): every threat likelihood and impact uses a WEP band (osint-tradecraft-standards.md §3.1); every cited intent / capability claim carries an Admiralty grade (§2); ≥1 alternative-hypothesis position (ACH or Red-Team, §4) is surfaced for any headline threat judgement.

coalition-dynamics

Purpose. Group cohesion scores and cross-party alliance pairs for the period's named votes, positions, and committee outputs.

EP MCP inputs. get_voting_records, analyze_coalition_dynamics, compare_political_groups, analyze_voting_patterns.

Required sections.

  1. Group roster — EPP, S&D, Renew, Greens/EFA, ECR, PfE, ESN, NI — seat count + current observed cohesion %.
  2. Alliance pair table — top 5 pairs by agreement rate this period, with trend vs. prior period.
  3. Defection highlights — named MEPs breaking with group line on named votes (with RCV IDs).
  4. Grand coalition status — viability indicator and named stress points.
  5. Confidence ledger — where roll-call data is available vs. where structural inference is used.

Mandatory Mermaid. graph LR linking groups with edge weight = agreement % (green >70%, orange 50-70%, red <50%).

Depth floor (breaking): 135 lines.

Mandatory SATs. ACH (§4.1 technique 1, primary — competing coalition hypotheses for each contested vote) · Indicators & Signposts (§4.1 technique 4, supporting — confidence ledger tripwires for when roll-call data publishes). See osint-tradecraft-standards.md §4.

Quality signals. At least one RCV ID cited; where EP voting feed has not yet published data, this is stated explicitly and group claims are marked LOW confidence.

Group-ID alias contract (MCP server v1.2.14+). The EP Open Data Portal exposes several groups under native-language acronyms that differ from the canonical English short codes used throughout analysis artifacts. As of european-parliament-mcp-server@1.2.15, the server canonicalises these aliases internally — do not add local remapping code in the calling workflow or in src/mcp/ep-mcp-client.ts. Pass the alias as received; the tool resolves it to the canonical ID before returning results. If the pinned server version is <1.2.14, the caller is responsible for translating aliases before calling the tool. Key alias mappings handled upstream (v1.2.14+):

Native alias(es) Canonical groupId
PPE, PPE-DE, Groupe du Parti populaire européen… EPP
SOC, PSE, Groupe de l'Alliance progressiste… S&D
Verts-ALE, Groupe des Verts/Alliance libre européenne Greens/EFA
ALDE, RE Renew
Native NI variants NI
GUE/NGL historical labels The Left

If a future run finds coverage.unrecognizedGroups is non-empty, update the upstream alias table in Hack23/European-Parliament-MCP-Server (src/tools/analyzeCoalitionDynamics.ts — POLITICAL_GROUP_ALIASES) rather than adding a local workaround.

cross-run-diff

Purpose. Bayesian delta vs. the previous run for the same article type: what changed in the data, what changed in the analytical assessment, and why.

EP MCP inputs. Reads the prior run's manifest.json + synthesis-summary + risk-matrix; optionally re-queries EP MCP only for fields that have refreshed.

Required sections.

  1. Prior-run header — date, run number, key findings, key scores.
  2. Data delta — new documents, new votes, new events since prior run.
  3. Assessment delta — each top-5 finding from prior run, with status (confirmed / upgraded / downgraded / reversed) and Bayesian posterior.
  4. Confidence migration — which claims moved between 🔴/🟡/🟢.
  5. Open questions — what the next run should investigate.

Mandatory Mermaid. flowchart LR with three columns: Prior → New Evidence → Posterior, one row per top finding.

Depth floor (breaking): 100 lines.

Mandatory SATs. Bayesian Update (§4.2 supporting, primary — prior × likelihood → posterior for each top finding) · Quality of Information Check (§4.1 technique 3, primary — confidence migration section). See osint-tradecraft-standards.md §4.

Quality signals. Prior-run artifact paths are valid; posterior assessments state the evidence class (supporting / contradicting / orthogonal) per political-risk-methodology.md §Bayesian Update. Source-grade × Estimative-language discipline (tradecraft): every posterior claim is expressed as a WEP band (osint-tradecraft-standards.md §3.1); evidence updating the prior carries an Admiralty grade (§2); primary SATs are Bayesian Update (§4.2) and Quality of Information Check (§4 technique 3).

economic-context

Purpose. Anchor the period's policy topics in IMF (sole authoritative source) macro, fiscal, trade, monetary, exchange-rate, banking-soundness, and commodity data.

EP MCP inputs. IMF native client (scripts/imf-mcp-probe.sh, tools imf-fetch-data / imf-search-databases / imf-get-parameter-defs / imf-get-parameter-codes) is the sole authoritative source for economic / fiscal / monetary / trade / FDI / exchange-rate / banking-soundness claims. World Bank MCP (worldbank-get-social-data, worldbank-get-health-data, worldbank-get-education-data) is used for non-economic context (health, education, social, environment, demographics, defence, agriculture, innovation, governance). See policy.

Required sections.

  1. Topic-to-indicator mapping — table linking each EP policy topic discussed this period to ≥1 IMF SDMX series (required for economic topics) plus optional WB non-economic cross-refs. See imf-indicator-mapping.md and (non-economic) worldbank-indicator-mapping.md.
  2. EU-27 headline indicators — IMF WEO latest values + 5-year trend + delta vs. EU average.
  3. Affected member-state focus — ≥3 member states most exposed to the period's dominant policy; compare IMF indicators.
  4. Forward outlook — IMF WEO / Fiscal Monitor projection data (+5y). Apply optimism-bias caveat sized per ../imf/forecast-accuracy-baseline.md for horizons ≥3y.
  5. Analytical bridge — how the data should shape the period's political reading.
  6. Data-source bridge — IMF vintage authoritative; WB non-economic cross-refs declared additive.
  7. SEO/evidence bridge — one short paragraph naming the search-intent policy terms (committee acronym, procedure title, affected stakeholder) that the article title/description may safely reuse from this artifact. Do not write separate marketing copy; write evidence-backed phrasing that can support manifest.title / manifest.description.

Mandatory Mermaid. One xyChart showing the headline IMF indicator over time, plus a flowchart LR mapping indicators to policy topics.

Depth floor (breaking): 185 lines.

Mandatory SATs. Quality of Information Check (§4.1 technique 3, primary — IMF source grading and data-vintage audit) · Bayesian Update (§4.2 supporting, supporting — forward-outlook section: IMF WEO projection vs prior-run vintage). See osint-tradecraft-standards.md §4.

Quality signals. Every IMF indicator cites its SDMX code + vintage; the enclosing HTML <section> carries data-vintage="WEO-April-2026"; every forecast number is within 30 words of a forecast marker (forecast/projection/projects/expects). Per-article-type IMF indicator floor (see imf-indicator-mapping.md §8) is satisfied. For Tier-1 articles citing high-sensitivity indicators, the cross-source triangulation outcome is logged per ../imf/cross-source-triangulation.md. The final bridge paragraph explicitly links the IMF signal to an EP political mechanism and gives safe SEO terms for title/description reuse.

historical-baseline

Purpose. Anchor every current score / metric in the run to its 30-day and 90-day historical baseline so the reader can tell rising from stable from declining.

EP MCP inputs. get_all_generated_stats, previous runs in analysis/daily/, monitor_legislative_pipeline trend, compare_political_groups over periods.

Required sections.

  1. Metric roster — table: metric, current value, 30-day avg, 90-day avg, trajectory arrow, confidence.
  2. Records and rarities — named "highest since …", "first occurrence in EP10", "return to baseline after …" findings.
  3. Cohort comparisons — current run vs. comparable runs earlier in the quarter / term.
  4. Anomaly highlights — any metric ≥2σ from its 90-day mean.
  5. Confidence caveats — where baselines are thin or unavailable.

Mandatory Mermaid. xyChart or timeline showing the leading metric over ≥90 days with current run highlighted.

Depth floor (breaking): 190 lines.

Mandatory SATs. Bayesian Update (§4.2 supporting, primary — anchor current scores to priors) · Key Assumptions Check (§4.1 technique 2, supporting — validate that 30-/90-day windows are long enough for the claim). See osint-tradecraft-standards.md §4.

Quality signals. Every baseline claim cites a prior run or an EP stats endpoint; "first occurrence" claims are justified against at least two prior periods.

mcp-reliability-audit

Purpose. Endpoint-by-endpoint record of EP MCP availability and data freshness during the run, plus any upstream issues filed on Hack23/European-Parliament-MCP-Server.

EP MCP inputs. Every endpoint the run attempted; the audit is a side-effect of the data-collection phase.

Required sections.

  1. Endpoint scoreboard — table: tool name, attempts, successes, 4xx/5xx/timeouts, data-age when available.
  2. Per-endpoint findings — one subsection per failing or degraded endpoint with: symptom, likely root cause, repro attempt, workaround used in this run.
  3. Upstream issues — list of linked issues on Hack23/European-Parliament-MCP-Server with title + URL.
  4. Data-source bridge — which alternative sources (direct endpoint fallback, prior-run cache, IMF primary economic, WB non-economic) compensated for failures.
  5. Reliability index — overall reliability score for this run (0–100) with breakdown.

Mandatory Mermaid. flowchart LR of endpoints with green (ok), orange (degraded), red (failed) nodes.

Depth floor (breaking): 385 lines (this is the deepest breaking-run artifact — MCP reliability is a first-class product concern).

Mandatory SATs. Quality of Information Check (§4.1 technique 3, primary — the artifact IS the per-source quality audit) · Red Team / Devil's Advocate (§4.1 technique 7, supporting — stress-test the feed reliability scores to avoid over-optimistic triage). See osint-tradecraft-standards.md §4.

Quality signals. Every degraded or failed endpoint has one of three explicit dispositions: a cited GitHub issue link, a triage-table citation recorded as documented behaviour — see §11 #N, or a clear note that it is a fileable new bug not yet linked; workarounds are reproducible from the text. Triage-gate compliance: every degraded/failed row has been looked up in .github/prompts/07-mcp-reference.md §11 Audit-Recurring Triage Table; items matching a 🟢 LIMITATION or 🔵 CALLING-PATTERN row are recorded as documented behaviour — see §11 #N and excluded from "Issues needing creation"; only 🔴 real-bug findings (or new symptoms not in the table) are filed. Denominator exclusion (§5 Reliability Index): 🟢/🔵-classified items are excluded from both numerator and denominator when computing the success-rate component — do not count OJQ-404 (get_meeting_foreseen_activities, §11 #7), events-feed slow-feed timeouts (🟡 SLOW_FEED_WARNING, §11 #8), or procedures-feed recess-mode responses (recessMode: true, §11 #5) against the reliability score.

significance-scoring

Purpose. 5-dimension composite significance score per candidate item with publish / withhold decision.

EP MCP inputs. Every event or document being scored; cross-referenced with get_all_generated_stats for historical comparability.

Required sections.

  1. Rubric recap — the five dimensions and their weights (from political-classification-guide.md §significance).
  2. Scoring table — one row per candidate item with per-dimension score + composite + decision (Publish / Hold / Withhold).
  3. Top-item narrative — ≥100 words on the highest-scored item explaining each dimension score.
  4. Threshold comparison — composite vs. the 30-day median score and the all-time top-5 scores.
  5. Decision audit — explicit reason for every Publish / Withhold.

Mandatory Mermaid. pie of dimension weights plus a small bar chart of candidate items by composite score.

Depth floor (breaking): 105 lines.

Mandatory SATs. Key Assumptions Check (§4.1 technique 2, primary — assumptions underlying the significance rubric weightings) · Competing Hypotheses Matrix (§4.1 technique 10, supporting — when two items compete for the same significance tier). See osint-tradecraft-standards.md §4.

Quality signals. Composite scores match the weighted sum of per-dimension scores; top item has evidence citations per dimension.

political-threat-landscape

Purpose. Threat Landscape view of the period applied to the six purpose-built EP democratic-threat dimensions defined in political-threat-framework.md. May live under intelligence/ or threat-assessment/ depending on the article type.

Required sections.

  1. Six-dimension scoreboard — each dimension scored 0–5 with a sentence of evidence.
  2. Top named threats — 3–5 with actor, mechanism, and affected institution.
  3. Institutional resilience — EP rules, treaty provisions, or institutional practices that offset each threat.
  4. Watchlist — concrete events to monitor.

Mandatory Mermaid. graph TD with a centre node "EP Democratic Integrity" and six colour-coded branches (one per dimension).

Depth floor (breaking): 90 lines (minimum); threat-heavy runs will routinely exceed this.

Mandatory SATs. Key Assumptions Check (§4.1 technique 2, primary) · Red Team / Devil's Advocate (§4.1 technique 7, primary — ≥1 alternative dimension-score for every headline threat) · Indicators & Signposts (§4.1 technique 4, supporting — watchlist section). See osint-tradecraft-standards.md §4.

Quality signals. Every named threat resolves to specific EP activity (procedure, MEP, group manoeuvre); no software-centric models are used. Source-grade × Estimative-language discipline (tradecraft): every dimension score (0–5) is accompanied by a WEP band for the underlying judgement (osint-tradecraft-standards.md §3.1); every cited threat actor claim carries an Admiralty grade (§2); primary SAT is Key Assumptions Check (§4 technique 1) with a Red-Team alternative (§4 technique 11).

wildcards-blackswans

Purpose. Low-probability / high-impact reserve watchlist — events that would invalidate the main scenarios if they occurred.

Required sections.

  1. Wildcard roster — ≥5 wildcards, each with probability (0–10%), impact (High / Severe), trigger signals.
  2. Quadrant chart — Probability × Impact placing each wildcard.
  3. Early-warning matrix — what signals would raise the wildcard to a scenario.
  4. Cascade map — which wildcard triggering would activate which other wildcards.

Mandatory Mermaid. quadrantChart with Probability (x) × Impact (y), every wildcard labelled. Black-swan zone (low prob, severe impact) in red.

Depth floor (breaking): 275 lines.

Mandatory SATs. High-Impact / Low-Probability Analysis (§4.1 technique 6, primary) · Indicators & Signposts (§4.1 technique 4, primary — early-warning matrix) · What-If Analysis (§4.1 technique 5, supporting — each wildcard activates a What-If chain). See osint-tradecraft-standards.md §4.

Quality signals. No wildcard is boilerplate ("economic crisis"); each is named to a plausible EP / EU trigger. Source-grade × Estimative-language discipline (tradecraft): wildcard probability is expressed as a WEP band (typically Very Unlikely / Remote Chance, osint-tradecraft-standards.md §3.1) with explicit time horizon (§3.4); trigger-signal sources carry Admiralty grades (§2); primary SATs are High-Impact / Low-Probability Analysis (§4 technique 10) and Indicators / Signposts (§4 technique 7).

reference-analysis-quality

Purpose. Self-score of this run against the reference benchmark (Run 184) plus a Pass-2 action list.

Required sections.

  1. Per-artifact line count vs. depth floor table.
  2. Per-artifact Mermaid presence + theme compliance check.
  3. Per-artifact evidence-density score.
  4. Benchmark gap narrative — ≥100 words identifying the run's weakest artifacts and why.
  5. Pass-2 action list — specific lines / sections to expand.

Mandatory Mermaid. flowchart LR Pass 1 → gap detected → Pass 2 → reference-quality exit.

Depth floor (breaking): 190 lines.

Mandatory SATs. Quality of Information Check (§4.1 technique 3, primary — the artifact IS a quality audit) · Key Assumptions Check (§4.1 technique 2, supporting — validate that benchmark comparisons are methodologically equivalent). See osint-tradecraft-standards.md §4.

Quality signals. Gap claims match the validator output; action list items are concrete enough that Pass 2 can execute them without re-reading the whole run.

voting-patterns

Purpose. Group-by-group bloc-behaviour analysis for the period — cohesion per group, observed coalitions, bloc win-rate, outlier votes, forward-vote forecasts. Distinct from coalition-dynamics (which is alliance-pair focused): voting-patterns answers "how did each of the 8 groups behave?"

EP MCP inputs. get_voting_records per plenary session covered, analyze_voting_patterns, analyze_coalition_dynamics, compare_political_groups, track_mep_attendance (for attendance-weighted cohesion).

Required sections.

  1. Group size & theoretical coalition arithmetic — ≥8 groups with seat counts, % of 720, one-line strategic role.
  2. Observed coalition patterns — table of ≥4 coalitions (Grand Centre, Progressive-Centrist, Conservative-Right, Opposition) with typical majority size, use-case policy domains, cohesion %, ≥1 RCV citation per row.
  3. Per-group behaviour — ≥7 groups each with internal cohesion %, dominant position, notable defections (named MEPs + RCV IDs), cross-group alliances.
  4. Bloc-behaviour index — per-bloc win-rate with trend vs. prior period.
  5. Stress points & outlier votes — ≥3 named RCVs where coalitions flipped, with ≥30-word explanation.
  6. Forward implications — ≥3 upcoming votes with expected coalition, confidence level, flip-conditions.

Mandatory Mermaid. graph LR of groups linked by agreement-rate edges; EPP blue, S&D red, Renew orange, Greens green, ECR light-blue (party colours aligned with the Hack23 semantic palette where possible).

Depth floor (breaking): 150 lines (weekly / other: 120 lines).

Mandatory SATs. ACH (§4.1 technique 1, primary — competing coalition hypotheses for each contested bloc vote) · Bayesian Update (§4.2 supporting, supporting — bloc win-rate trend vs. prior period). See osint-tradecraft-standards.md §4.

Quality signals. ≥5 RCV IDs cited; every cohesion % has a vote-count backing; aggregate-only claims explicitly flagged LOW confidence where EP roll-call data has not yet published. §"Voting Data Freshness" table present with source field populated (mcp / ep-open-data-portal / unavailable); when source is ep-open-data-portal the CC BY 4.0 attribution string is present and WEP bands widened +5 pp; when source is unavailable all coalition claims carry explicit LOW-confidence flag and WEP bands widened +10 pp per osint-tradecraft-standards.md §3.1. The D-02 fallback path (getVotingRecordsWithFallback in src/mcp/ep-open-data-client.ts) is invoked whenever get_voting_records returns an empty votes array — structural-proxy cohesion scores may never be used as a silent substitute for real voting data.

workflow-audit

Purpose. End-of-run self-audit of workflow execution — phases completed, MCP tools called, Core Principles compliance, time budget, issues and deviations. Produced at Step 10 of the 10-step protocol by every news-* workflow.

EP MCP inputs. None directly. Reads the workflow log, the run's data/ directory, the run's artifact filesystem, and manifest.json.

Required sections.

  1. Metadata YAML block — articleType, runId, date, confidenceLevel, rulesAudited, complianceRate.
  2. Workflow execution summary — 6-phase table with status per phase (Health gate → Data collection → Editorial context → Analysis → Significance gate → Validation → PR creation).
  3. MCP tool call log — one row per MCP call with tool, args (truncated), result, record count, latency.
  4. Core Principles compliance — 11-row scorecard against ai-driven-analysis-guide.md §Core Principles, with evidence cell per row.
  5. Artifact production — folder-by-folder table (planned / produced / short-of-floor).
  6. Time budget — step-by-step target vs. actual.
  7. Issues & deviations — per-issue narrative (symptom / root cause / workaround / next-run recommendation).
  8. Recommendations for the next same-type run — ≥3 concrete items.

Mandatory Mermaid. flowchart LR of the 6 phases colour-coded by status (green=ok, orange=degraded, red=failed).

Depth floor (breaking): 100 lines (same across article types — this is a run-transparency artifact, not a content artifact).

Quality signals. Every ❌ or ⚠️ status has a named symptom; every next-run recommendation is specific enough that the next run can execute it without re-deriving it.

methodology-reflection

Purpose. Analytic-quality retrospective that closes the continuous-improvement loop. Distinct from workflow-audit (which checks mechanical phase-by-phase compliance): methodology-reflection.md reflects on which Structured Analytic Techniques (SATs) were applied, where biases were mitigated, what the evidence would not support, and what the next same-type run must do differently. Modelled on the Hack23 riksdagsmonitor interpellations reflection.

EP MCP inputs. None directly — built from the completed run artifacts, the workflow-audit.md, the MCP tool call log, and the editorial-review outcome (if any).

Required sections.

  1. Pipeline overview — colour-coded Mermaid graph TD from trigger → MCP pulls → Pass 1 → Pass 2 → significance gate → validation → PR → human editorial review → optional Pass 3.
  2. Data sources and provenance — ≥6 rows (tool, purpose, status ✅/⚠️/❌, confidence grade 🟢/🟡/🔴). Null-response signals ("returned 0 results, confirming no activity") captured explicitly.
  3. Structured Analytic Techniques applied — ≥10 SATs named (Classification, Significance scoring, SWOT, Risk matrix, Threat analysis, Stakeholder mapping, PESTLE, Scenario analysis, ACH, Key Assumptions Check, Red Team, Historical baseline, Bayesian cross-run delta, Comparative international, Per-document deep dives). Every SAT linked to an artifact.
  4. AI-FIRST iteration log — Pass 1 (count of artifacts, coverage gaps, depth gaps, SATs missing) + Pass 2 (concrete changes, word/citation counts added, remaining gaps) + optional Pass 3 (trigger + actions taken).
  5. Strengths — ≥5 with ≥1 specific citation each (evidence class, quantitative anchoring, pattern detection, SATs breadth, comparative context, confidence grading discipline).
  6. Limitations and caveats — ≥5 with honest specificity (MCP endpoint failures, data-freshness limits, single-run inference constraints, polling data unavailable, language / cultural bias).
  7. Lessons for future same-type runs — ≥5 concrete enough that the next run can execute them (data-collection lesson, SAT-timing lesson, comparative-context lesson, methodology-artifact lesson, iteration-time lesson).
  8. Known biases and mitigations — ≥6-row table (Confirmation · Availability · Mirror-imaging · Narrative fallacy · Recency · Selection · Anchoring), each with risk level, mitigation applied, and residual risk.
  9. Peer review / editorial oversight — per AI_Policy.md; name reviewer or state "pending editorial queue".
  10. Update plan — ≥5 triggers mapped to specific artifacts with frequency.
  11. References — Heuer (1999), Heuer & Pherson (2020), UK MoD Red Teaming Handbook, NATO AJP-2.1, Hack23 AI_Policy, .github/skills/ai-first-quality.md, plus run-specific references.

Mandatory Mermaid. graph TD pipeline diagram in §1 with the seven-colour Hack23 palette applied to node states (blue=input, green=safe-pass, orange=degraded-fallback, red=failed, purple=synthesis, yellow=gate, light-blue=human-review).

Depth floor (breaking): 220 lines. Other article types: 180 lines.

Quality signals. Every SAT row cites an artifact that actually exists in this run (or states "not-applicable because …"); every limitation is specific enough that a reader knows what could be wrong with the analysis; every lesson is actionable by the next run without re-derivation; bias mitigations reference named artifacts; peer-review status is honestly reported (never marked "reviewed" when it is pending).

Workflow placement. Produced as the final artifact of every run, after workflow-audit.md and after the PR patch is otherwise complete. Every news-*.md workflow (except news-translate.md) must write this file before calling safeoutputs___create_pull_request.

cross-session-intelligence

Purpose. Session-over-session narrative of parliamentary politics across plenary sessions within a period. Distinct from cross-run-diff (which is cross-run of the same article type) — this file tells the story of how the political programme matured across sessions.

EP MCP inputs. get_plenary_sessions (year-filtered), get_meeting_decisions per session ID, get_meeting_activities, get_adopted_texts (period), get_voting_records per session.

Required sections.

  1. Session overview — table with ≥2 sessions covering dates, sitting days, location, texts adopted, theme.
  2. Progression diagram — Mermaid timeline showing the period's sessions as a horizontal sequence with key themes.
  3. Session-by-session progression — ≥200 words per session covering character, political arc, ≥3 adopted-text IDs, rapporteurs.
  4. Cross-session themes — ≥4 themes with "sessions touched" columns and trajectory narrative (≥30 words each).
  5. Crystallisation moment — identify the period's single most strategically concentrated session with ≥250-word analysis.
  6. Momentum indicators — multi-metric table (texts/day, cohesion %, RCVs/day, attendance %) across all sessions with trend arrows.
  7. Forward outlook for next session — ≥3 topic forecasts with confidence levels.

Mandatory Mermaid. timeline of the period's sessions. Optional secondary flowchart LR showing cross-session theme threads.

Depth floor (breaking): 220 lines (motions quarterly-scope runs), 150 lines (week-in-review / month-in-review).

Mandatory SATs. Bayesian Update (§4.2 supporting, primary — session-over-session narrative tracks how assessments evolved) · Indicators & Signposts (§4.1 technique 4, supporting — forward outlook section). See osint-tradecraft-standards.md §4.

Quality signals. Sessions reference specific part-session IDs or date ranges; the crystallisation-moment section stands on its own as analysis (not a bullet list).

🏷️ classification/

significance-classification

Purpose. 5-dimension significance rubric per event + publish / withhold decision. Often identical in content to intelligence/significance-scoring.md but lives here when classification comes before intelligence synthesis in the run.

Inputs, sections, Mermaid, floor. Use the significance-scoring rules above.

actor-mapping

Purpose. Named actors with influence weights, committee seats, roll-call alignment patterns, and alliance footprints.

EP MCP inputs. get_meps, get_committee_info, assess_mep_influence, analyze_country_delegation, analyze_voting_patterns.

Required sections.

  1. Actor table — ≥12 named actors with role, institutional base, influence weight (0–10, justified).
  2. Alliance footprint — which actors reliably vote / speak together.
  3. Dissent footprint — which actors break with their group most often and on what topics.
  4. New-actor spotlight — any MEPs / groups who gained or lost influence this period.

Mandatory Mermaid. graph LR network of actors with edge weight = co-alignment rate.

Depth floor (breaking): flat 30 lines (not yet benchmarked).

Mandatory SATs. Stakeholder Mapping (§4.2 supporting, primary) · ACH (§4.1 technique 1, supporting — when ≥2 plausible alliance footprints compete). See osint-tradecraft-standards.md §4.

Quality signals. Every influence weight cites at least one EP MCP data source.

forces-analysis

Purpose. Driving forces vs. restraining forces on the period's dominant issue (Lewin force-field analysis adapted to EP politics).

Required sections.

  1. Issue frame.
  2. Driving forces — ≥5, each with magnitude 1–5 and origin (institutional / political / economic / external).
  3. Restraining forces — ≥5, same structure.
  4. Net pressure — forces summed, with direction and narrative.
  5. Intervention points — where a small input could flip the balance.

Mandatory Mermaid. flowchart LR with arrows sized by magnitude, driving forces green, restraining red, issue node in centre.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. Force-Field Analysis (§4.2 supporting, primary — driving vs. restraining forces IS the SAT) · Key Assumptions Check (§4.1 technique 2, supporting — validate force magnitude scores). See osint-tradecraft-standards.md §4.

Quality signals. Every force is named to a specific EP, EU, or geopolitical source.

impact-matrix

Purpose. Event × stakeholder × dimension matrix showing who is affected by what, and how.

Required sections.

  1. Event list — ≥5 named events / decisions from the period.
  2. Stakeholder list — ≥6 named stakeholder groups.
  3. Matrix — rows = events, columns = stakeholders, cells = impact class (🟢 Positive · 🟡 Mixed · 🔴 Negative · ⚪ None) with one-line justification.
  4. Hot-cell narratives — ≥3 paragraphs expanding the most consequential cells.

Mandatory Mermaid. None required (matrix is a table); optional flowchart LR event → stakeholder if the chain is dense.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. Stakeholder Mapping (§4.2 supporting, primary — the matrix grid IS the stakeholder impact map) · What-If Analysis (§4.1 technique 5, supporting — cascade cells require a "what if this hits" chain). See osint-tradecraft-standards.md §4.

Quality signals. No cell is unjustified; hot-cell narratives cite specific procedures or documents.

⚠️ risk-scoring/

risk-matrix

Purpose. 5×5 Likelihood × Impact matrix with ≥5 named political risks and monitoring triggers. Canonical methodology: political-risk-methodology.md.

Required sections.

  1. Risk roster — ≥5 named risks with L (1–5), I (1–5), score (L×I), category (Coalition / Policy / Institutional / External / Data).
  2. 5×5 matrix — Mermaid quadrantChart placing each risk.
  3. Per-risk narrative — ≥100 words per top-3 risk with evidence, stakeholder exposure, and monitoring trigger.
  4. Trend vs. prior run — each risk scored last time + delta.
  5. Accept / Prepare / Monitor decision — per risk.

Mandatory Mermaid. quadrantChart 5×5 with labels Low/High on both axes, monitored-zone green, elevated-zone orange, critical-zone red.

Depth floor (breaking): 150 lines.

Mandatory SATs. Key Assumptions Check (§4.1 technique 2, primary — assumptions underlying L and I scores) · ACH (§4.1 technique 1, supporting — when multiple risks share monitoring triggers) · What-If Analysis (§4.1 technique 5, supporting — tail-risk entries). See osint-tradecraft-standards.md §4.

Quality signals. Every score matches L × I; every monitoring trigger is time-bounded. Source-grade × Estimative-language discipline (tradecraft): each risk narrative reports the residual likelihood as a WEP band (osint-tradecraft-standards.md §3.1) with time horizon (§3.4); evidence backing the L and I scores carries an Admiralty grade (§2); ACH (§4 technique 6) is used when multiple competing risks share the same monitoring triggers.

📎 Worked example: examples/risk-matrix-good-output.md — illustrative 5-risk register with WEP-banded Likelihood, dated trigger events, named mitigation owners, and pre/post-mitigation residual scoring.

quantitative-swot

Purpose. 3+3+3+3 SWOT with numeric weights plus TOWS cross-quadrant strategies. Canonical methodology: political-swot-framework.md.

Required sections.

  1. Quadrants — Strengths, Weaknesses, Opportunities, Threats — each ≥3 items, each item with severity (high/medium/low) + ≥80 words of evidence.
  2. TOWS matrix — SO / ST / WO / WT strategies, one per cell.
  3. Cross-quadrant interference — which S reinforces which O, which W amplifies which T.
  4. Scenario bridge — which SWOT configuration points to which scenario in scenario-forecast.md.

Mandatory Mermaid. quadrantChart with axes External/Internal × Positive/Negative; optional graph LR for TOWS bridges.

Depth floor (breaking): 140 lines.

Mandatory SATs. SWOT (Heuer & Pherson SWOT framework, primary — the four quadrants are the SAT output) · Bayesian Update (§4.2 supporting, supporting — TOWS scenario bridge section maps SWOT configuration to scenario-forecast posteriors). See osint-tradecraft-standards.md §4.

Quality signals. Every SWOT item has an evidence citation; no quadrant has fewer than 3 items; TOWS strategies name specific actors and timelines.

📎 Worked example: examples/swot-good-output.md — illustrative excerpt showing ≥80-word cells with evidence + confidence labels and four TOWS pairs with dated triggers and named owners.

political-capital-risk

Purpose. Named rapporteur / chair / group-leader political capital exposure per top position in the period.

Required sections.

  1. Named actor capital table — actor, position, capital invested, counter-party, downside if defeated.
  2. Capital flow diagram — where capital is being spent and what payoff is expected.
  3. Precedent table — similar bets from the past and their outcomes.
  4. Outlook — capital likely to be preserved / eroded / lost.

Mandatory Mermaid. graph LR capital flow from actor to procedure to outcome.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. Stakeholder Mapping (§4.2 supporting, primary — capital flow is a form of stakeholder power dynamics) · Key Assumptions Check (§4.1 technique 2, supporting — assumptions about what constitutes "political capital" for each actor). See osint-tradecraft-standards.md §4.

Quality signals. Every named actor has an EP role cited.

legislative-velocity-risk

Purpose. Throughput, stalled procedures, and deadline exposure vs. term end. Canonical upstream methodology: political-risk-methodology.md §Velocity.

Required sections.

  1. Pipeline summary — open procedures by stage, throughput per week, median time-in-stage.
  2. Stalled procedures — ≥3 named, with time-in-stage, likely cause, and rescue paths.
  3. Deadline exposure — procedures at risk of expiring before term end.
  4. Bottleneck map — where the process is slowest and why.

Mandatory Mermaid. gantt of top procedures' remaining stages, or flowchart LR bottleneck map.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. Indicators & Signposts (§4.1 technique 4, primary — deadline exposure section and bottleneck map are indicator-based) · Key Assumptions Check (§4.1 technique 2, supporting — assumptions about term-end deadlines). See osint-tradecraft-standards.md §4.

Quality signals. Uses monitor_legislative_pipeline data explicitly.

🎭 threat-assessment/

consequence-trees

Purpose. Consequence tree per named threat: action → first-order → second-order → democratic outcome.

Required sections.

  1. Threat roster — ≥3 named threats.
  2. Per-threat consequence tree — ≥3 levels deep with probability / severity per branch.
  3. Cross-tree convergence — where multiple threats hit the same democratic outcome.
  4. Intervention points — where the tree can be pruned by EP / EU action.

Mandatory Mermaid. graph TD tree per threat, red root → orange mid → yellow leaves.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. What-If Analysis (§4.1 technique 5, primary — "what if this threat action fires?" drives each tree branch) · High-Impact / Low-Probability Analysis (§4.1 technique 6, supporting — cross-tree convergence section focuses on low-probability, high-impact combinations). See osint-tradecraft-standards.md §4.

Quality signals. No branch is "and then bad things happen"; every branch names a mechanism.

legislative-disruption

Purpose. How adversarial pressure could stall, redirect, or capture specific procedures.

Required sections.

  1. Targeted-procedure list — ≥3 procedures with procedure ID, rapporteur, current stage, disruption opportunity score.
  2. Disruption playbook — per procedure, named techniques (amendment flooding, rapporteur targeting, committee obstruction, trilogue stalling).
  3. Detection indicators — what would reveal disruption early.
  4. Institutional counter-measures — EP Rule of Procedure tools available.

Mandatory Mermaid. flowchart LR procedure → disruptor vector → detection → counter.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. What-If Analysis (§4.1 technique 5, primary — "what if this disruption technique is applied?" drives the disruption playbook) · Red Team / Devil's Advocate (§4.1 technique 7, supporting — the disruption playbook is itself a Red Team exercise). See osint-tradecraft-standards.md §4.

Quality signals. Every technique is linked to a named EP Rule of Procedure or precedent.

actor-threat-profiles

Purpose. Per-threat-actor profile using the Diamond Model (adversary / capability / infrastructure / victim) adapted for political actors.

Required sections.

  1. Actor roster — ≥3 named actors.
  2. Per-actor Diamond — intent, capability, opportunity, attack surface.
  3. Relationship map — how actors cooperate, compete, offset each other.
  4. Escalation paths — how each profile could escalate in severity.

Mandatory Mermaid. graph LR actors × target institutions with edge labels for capability.

Depth floor (breaking): flat 30 lines.

Mandatory SATs. Red Team / Devil's Advocate (§4.1 technique 7, primary — the Diamond Model is applied from the adversary's perspective) · Key Assumptions Check (§4.1 technique 2, supporting — validate intent and capability assumptions for each actor). See osint-tradecraft-standards.md §4.

Quality signals. No personal-life data on MEPs; analysis stays strictly on public parliamentary role.

threat-analysis

Purpose. Tabular per-threat record — one row per identified democratic-process threat with vector × likelihood (WEP) × impact × mitigation × owner. Produced when the four landscape artifacts (political-threat-landscape, actor-threat-profiles, consequence-trees, legislative-disruption) need a structured roll-up an executive can scan in 60 seconds. Uses the framework template threat-analysis.md.

Required sections.

  1. BLUF — single-paragraph judgment of the period’s top threat by Likelihood×Impact, with WEP band and Admiralty grade on the underlying evidence.
  2. Threat register — one row per threat with columns: ID, vector, target dimension (one of the 6 Political Threat Landscape dimensions), Likelihood (WEP), Impact (1–5), composite score, mitigation, owner, evidence (Admiralty + procedure / RCV reference).
  3. Top-3 threats deep-dive — one paragraph each, naming the kill-chain stage at which the threat is interruptible.
  4. Cross-references — ≥1 link into political-threat-landscape.md and ≥1 link into consequence-trees.md so the reader can pivot from the table into the qualitative analysis.
  5. Tradecraft footer — ≥1 SAT named (e.g. Indicators & Signposts, Pre-Mortem); STRIDE explicitly NOT used.

Mandatory Mermaid. graph TD — vector × mitigation matrix; or kill-chain phase × interruption point.

Depth floor. Flat 30 lines (matches the four sister landscape artifacts).

Quality signals.

  • Every Likelihood is a WEP band (not a bare number); every Impact is justified by one observable consequence.
  • Mitigation owners are named institutionally (e.g. “ENVI chair”, “EPP whip”), never by personal name unless the role is the relevant unit of analysis.
  • The table totals match the BLUF — if BLUF says “top threat is coalition fracture”, that row’s composite score must be the table maximum.

What good looks like.

BLUF: The dominant threat across the 21–25 April window is a Grand Coalition fracture on IED enforcement provisions — LIKELY to crystallise at the 12 May trilogue (WEP: LIKELY 65–85%, evidence A2 from RCV-2026-0412). Mitigation owner: EPP whip’s office. Kill-chain interruption point: pre-trilogue shadows meeting (8 May).

📄 documents/

document-analysis-index

Purpose. One row per downloaded EP document with its per-file analysis path, status, classification, significance score, and link to the raw MCP data file.

Required sections.

  1. Feed-by-feed download summary — how many items per feed succeeded / failed.
  2. Document table — one row per document: id, type, title (first 80 chars), feed, analysis file path, status, significance score.
  3. Coverage metrics — % of downloaded documents with analysis file; top gaps.
  4. Links to representative deep-dives — ≥3 high-significance per-file analyses.

Mandatory Mermaid. flowchart LR feed → raw-data/ → per-file analysis.md → aggregation into intelligence/.

Depth floor (breaking): 95 lines.

Quality signals. Every analysis file path in the table resolves to an existing file with ≥100 lines.

📜 existing/ (legacy long-form layout)

The existing/ folder is the canonical location for long-form prose artifacts used by motions, propositions, and quarter / month-in-review workflows. It is a layout convention, not a new artifact group — files here co-exist with the 5 standard folders (intelligence/, classification/, risk-scoring/, threat-assessment/, documents/) and carry content the standard folders cannot accommodate (10 000-word political prose; multi-session fact rosters). Newer runs may place the same content under intelligence/ instead; both locations are valid.

deep-analysis

Purpose. Long-form (4 000–10 000 word) Economist-style political intelligence prose — the 30-minute read complement to the 5-minute synthesis-summary.md. Primary output of motions-run* workflows, secondary output of month-in-review and propositions workflows.

EP MCP inputs. Consumes the run's session-baseline.md, voting-patterns.md, cross-session-intelligence.md, coalition-dynamics.md, stakeholder-map.md. Pulls named adopted-text IDs directly from get_adopted_texts results in data/.

Required sections.

  1. Executive summary — ≥200 words stating the political thesis of the period.
  2. Structural thesis — ≥400 words identifying ≥3 named policy clusters with ≥2 named texts per cluster, coalition footprint, historical comparison.
  3. Crystallisation moment — ≥500 words on the period's defining session with architectural, procedural, and coalition angles; ≥3 adopted-text IDs and ≥2 RCV IDs cited inline.
  4. Coalition dynamics — ≥500 words on the Grand Centre (or dominant coalition) performance with ≥2 named MEPs whose defections mattered.
  5. Policy dimensions — ≥4 sub-sections, ≥300 words each (Trade / Defence / Digital / Environment / etc.), each with committee, rapporteur, adopted texts, coalition behaviour.
  6. Institutional dynamics — ≥400 words on the EP ↔ Commission ↔ Council triangle.
  7. Geopolitical context — ≥400 words grounding the period in external events and macro indicators.
  8. Forward trajectory — ≥400 words with ≥5 forward monitors and ≥2 probability-weighted scenarios.
  9. Confidence & method note — ≥150 words explaining data-source bridges, gaps, and what would change the assessment.

Mandatory Mermaid. ≥3 diagrams across the file (structural thesis diagram + coalition network + forward trajectory flowchart are typical).

Depth floor (breaking): 400 lines (motions / month-in-review), 200 lines (week-in-review).

Quality signals. ≥15 named procedures / adopted texts / RCVs cited inline; ≥2 historical comparisons (prior term / prior year); named coalitions with explicit cohesion percentages; no partisan conclusions.

session-baseline

Purpose. Structured calendar + adopted-texts roster for every plenary session in scope. The data-dense reference that other artifacts cite ("see session-baseline §4.2"). Distinct from historical-baseline (metric trending) — session-baseline is the calendar.

EP MCP inputs. get_plenary_sessions (year-filtered), get_adopted_texts (period), get_procedures, get_committee_info, track_mep_attendance for attendance means.

Required sections.

  1. Run context — date, run ID, analysis directory, article type, parliament term, period covered.
  2. Plenary session calendar — one sub-section per session in scope with dates, location, sitting days, texts adopted, RCVs, key theme, EP session ID.
  3. Session calendar diagram — Mermaid gantt with Strasbourg vs. Brussels sections and each session as a bar.
  4. Period totals — sitting days, texts, RCVs, resolutions (non-binding), legislative acts (binding), mean attendance.
  5. Adopted texts roster — per-session table (≥1 row per adopted text in scope) with adopted-text ID, title (first 80 chars), procedure code, committee, domain.
  6. Committee activity map — ≥5 committees with texts reported, rapporteurships, shadow rapporteurships, dominant group.
  7. Procedure-code distribution — count and % per procedure type (COD / CNS / APP / INI / Budget / Other).
  8. Historical anchor — current period vs. same-quarter-prior-year vs. prior-quarter vs. term-to-date average with ≥100-word narrative.
  9. Data-source ledger — ≥5 MCP tools with records fetched, status, notes.

Mandatory Mermaid. gantt of the session calendar.

Depth floor (breaking): 200 lines (motions / month-in-review), 150 lines (weekly / breaking).

Quality signals. Every session row has an explicit EP session ID or an explicit "session ID unavailable via MCP this run" note; every adopted text has a TA-YY-YYYY-NNNN ID.

Mirror artifacts in existing/

Older motions-* runs mirror intelligence/coalition-dynamics.md, intelligence/stakeholder-impact.md, and intelligence/synthesis-summary.md into existing/ alongside deep-analysis.md and session-baseline.md. Construction rules for these mirrors are identical to their intelligence/ counterparts — follow the intelligence/ sections above. The duplication is a legacy layout side-effect; newer runs write only to the intelligence/ location.

🧭 extended/ — Optional Deep-Intelligence Artifacts

Scope. The 12 artifacts below were introduced by the riksdagsmonitor port (2026-04-23). They are optional (not gated by the completeness-gate) but recommended for long-form review workflows and crisis / breaking deep runs. Construction rules follow the same 2-pass AI-First pattern (Pass 1 draft → Pass 2 read-back and improve) as the mandatory artifacts.

executive-brief

Purpose. ≤90-second decision brief for senior readers: BLUF + 3–5 key judgments (each with confidence label and source Admiralty grade) + 3 recommended actions with monitoring windows. Derived from synthesis-summary.md and the top-ranked significance-scoring.md items.

Required sources. intelligence/synthesis-summary.md (read entirely), intelligence/significance-scoring.md (top 3 items), extended/intelligence-assessment.md if present.

Construction steps. (1) Extract top-3 significance items. (2) Write BLUF (25–50 words). (3) Derive 3–5 key judgments with WEP bands and confidence levels. (4) Specify 3 actions with named owners and monitoring dates. (5) Admiralty grade every claim.

Quality signals. ≥150 lines. ≥5 Admiralty-graded evidence cells. All judgments WEP-labelled. Every action has an owner + monitoring date.

Cross-refs. → intelligence-assessment.md · → synthesis-summary.md.

devils-advocate-analysis

Purpose. Adversarial challenge to the run's consensus view. Applies ACH (Analysis of Competing Hypotheses) and Red Team SAT.

Required sources. intelligence/synthesis-summary.md, intelligence/scenario-forecast.md, plus the primary MCP evidence. analyze_coalition_dynamics, detect_voting_anomalies, get_voting_records.

Construction steps. (1) State the consensus view. (2) List 2–4 competing hypotheses. (3) Build an ACH matrix (hypothesis × evidence, consistency markings ++/+/0/−/−−). (4) Red Team attack each consensus assumption. (5) Conclude with confidence-updated assessment.

Quality signals. ≥240 lines. ACH matrix with ≥8 evidence rows. ≥3 competing hypotheses. WEP band per hypothesis. Explicit "what would change my mind" section.

Cross-refs. → osint-tradecraft-standards.md §ACH · → scenario-forecast.md.

historical-parallels

Purpose. Map ≥3 EU historical precedents with parallel mechanisms and divergence points. EU-specific precedents: Santer Commission resignation (1999), Article 7 TEU proceedings vs Poland (2017–) and Hungary (2018–), Brexit triggering (2017), MFF 2014–2020 negotiations, eurozone crisis (2010–2012), Constitutional Treaty referendum failures (2005), Commission-censure motions.

Required sources. EP Open Data historical document registry, get_adopted_texts (historical years), search_documents, externally verified historical facts.

Construction steps. (1) Name 3–5 precedents. (2) For each: mechanism, outcome, timeline. (3) Map parallels to current situation (≥3 mechanism matches per precedent). (4) Identify divergence points. (5) Base-rate calculation.

Quality signals. ≥230 lines. ≥3 precedents with ≥5 mechanism-match cells each. Explicit base-rate estimates with CI. Divergence section ≥100 words.

Cross-refs. → historical-baseline.md · → scenario-forecast.md.

coalition-mathematics

Purpose. Vote-math analysis for the period's contested files. Uses post-2024 EP10 seat counts: 720 seats total, 361 absolute majority. Grand coalition (EPP 188 + S&D 136 + Renew 77 = 401); Left-of-centre (S&D 136 + Renew 77 + Greens 53 + Left 46 = 312); Right-of-centre (EPP 188 + ECR 78 + PfE 84 = 350).

Required sources. get_voting_records, analyze_voting_patterns, analyze_coalition_dynamics, compare_political_groups, analyze_country_delegation.

Construction steps. (1) List contested files. (2) For each, compute pass/fail margins across standard coalitions. (3) Identify swing groups. (4) Model national-delegation defection scenarios (top 5: DE 96, FR 81, IT 76, ES 61, PL 53). (5) Project forward votes with WEP bands.

Quality signals. ≥280 lines. ≥3 coalition permutations per file. Swing-group identified. National-delegation defection modelled for ≥2 scenarios. Mermaid flowchart showing arithmetic.

Cross-refs. → coalition-dynamics.md · → voting-patterns.md · → synthesis-methodology.md.

forward-indicators

Purpose. Leading-indicator watchlist with WEP bands and monitoring cadence. Applies Indicators & Signposts SAT.

Required sources. monitor_legislative_pipeline, track_mep_attendance, get_plenary_sessions, get_events, detect_voting_anomalies, Eurobarometer.

Construction steps. (1) Enumerate 8–15 indicators across coalition stability, legislative throughput, institutional integrity, electoral signals. (2) Per indicator: trigger + detection + cadence + threshold. (3) Dashboard table. (4) Mermaid flowchart of alerting pipeline.

Quality signals. ≥250 lines. ≥8 indicators. Every indicator has a specific MCP tool or external source. Escalation thresholds numeric.

Cross-refs. → osint-tradecraft-standards.md §Indicators & Signposts SAT · → scenario-forecast.md · → significance-scoring.md.

intelligence-assessment

Purpose. ICD 203 compliant IC-style assessment. BLUF, Key Judgments (with confidence + WEP), Discussion, Outlook, Alternative Hypotheses, Gaps / Caveats, Sources.

Required sources. Every artifact in intelligence/ and classification/ for the run.

Construction steps. (1) BLUF (40–80 words). (2) 3–7 Key Judgments with ICD 203 Standard 8 confidence labels. (3) Discussion with evidence citations (Standard 6). (4) Outlook — WEP-labelled 30/90-day projection. (5) Alt hypotheses. (6) Information gaps. (7) Source Admiralty table.

Quality signals. ≥270 lines. ≥3 key judgments. ICD 203 Standards 2/4/6/8 demonstrably present. Source table with ≥10 graded sources.

Cross-refs. → executive-brief.md · → osint-tradecraft-standards.md §ICD 203 Mapping.

implementation-feasibility

Purpose. 27 Member State implementation feasibility across 5 dimensions: legal transposition complexity, administrative capacity, fiscal cost, compliance monitoring, enforcement risk. Heatmap + MS-by-MS risk register.

Required sources. track_legislation, get_procedures, get_external_documents (Council + Commission IA), IMF GFS + FM fiscal data (primary for any fiscal / budgetary feasibility claim), World Bank governance indicators (WGI — non-economic).

Construction steps. (1) Extract concrete obligations from proposal. (2) Score each MS × 5 dimensions (1–5). (3) Heatmap. (4) Flag high-risk MS (≥3 dimensions scoring ≥4). (5) Identify mitigations and flexibility mechanisms.

Quality signals. ≥290 lines. 27 MS × 5 dimensions = 135 cells filled. High-risk MS register with specific mitigations.

Cross-refs. → legislative-velocity-risk.md · → economic-context.md.

media-framing-analysis

Purpose. Comprehensive media framing & influence-operations analysis across EU-wide and national press. Identifies dominant narrative frames, manipulation vectors, cognitive vulnerabilities, narrative-laundering chains, and counter-resilience gaps using the Entman (1993), DISARM, and ABCDE (Camille François) frameworks. Covers ≥15 outlets across ≥5 MS/languages including Brussels bureaux (Politico EU, Euractiv, EUobserver), wire services (AFP, Reuters, DPA), national quality press (FAZ, Le Monde, Corriere, El País), and state-affiliated monitoring (RT, CGTN). Mandatory for every article-generating workflow (registered in src/config/article-horizons.ts mandatoryArtifacts for every slug).

When to build it. Pass 2 (preferred) or late Pass 1 — never first. The artifact depends on synthesis-summary, stakeholder-map, scenario-forecast, coalition-dynamics, and (for review runs) voting-patterns being stable so the framing analysis can cite them and identify which narratives amplify or contradict the underlying parliamentary record. Writing it earlier forces a Pass-2 rewrite when synthesis evidence shifts. See analytical-supplementary-methodology.md §AS4 → "When to Build It".

Required sources. External media (open-access only — no paywall bypass). EP MCP tools: get_speeches, get_adopted_texts, get_parliamentary_questions. Brussels bureaux (Admiralty B2), national quality press (Admiralty C2), wire services (Admiralty B2). State-affiliated outlets monitored for Frame E detection but never cited as factual authority (ceiling D3).

Construction steps. (1) Populate Tradecraft Context (F3EAD, PIRs, SATs, ICD 203). (2) Identify 3–5 frames and produce Frame Package Overview Mermaid. (3) Apply Entman functions (Problem / Cause / Moral / Treatment) per frame. (4) Map cognitive vulnerabilities per frame with primary academic literature. (5) Scan for DISARM TTPs using T-codes. (6) Trace narrative-laundering chain (fringe → alt-media → political amplification → mainstream → international). (7) Populate Source Ecology / Outlet Bias Audit with 5-axis multi-dimensional assessment (no-neutral-media doctrine). (8) Assess CIB indicators using ABCDE framework. (9) Document algorithmic-amplification asymmetry per platform with academic citations. (10) Identify ≥2 comparative-international frame cognates per frame. (11) Pattern-match against ≥5 strategic doctrines. (12) Produce frame lifecycle xychart (≥8 time points) with half-life and zombie probability. (13) Calculate RRPA composite (Reach × Resonance × Persistence × Action) with dated real-world evidence. (14) Populate counter-resilience plan (L1–L5). (15) Document quote salience with manipulation flags. (16) Produce frame-competition dynamics Mermaid. (17) Build coverage-volume dashboard (day-by-day). (18) EU vs national framing comparison (≥5 MS). (19) Forward watchlist with WEP + Admiralty. (20) Complete Pass-2 Self-Audit Checklist (33 items).

Quality signals. ≥270 lines (breaking baseline; per-article-type floors in reference-quality-thresholds.json scale from 180 for short-horizon to 320 for election-cycle). ≥15 outlets assessed across ≥5 MS/languages. All 4 Entman functions per frame. DISARM TTPs with verbatim T-codes. Narrative-laundering chain ≥4 stages. CIB ABCDE all 7 indicators assessed. ≥3 academic citations for algorithmic-amplification. ≥2 international cognates per frame (naivety check). Frame lifecycle xychart ≥8 time points. RRPA composite with dated indicators. Counter-resilience L1–L5 populated. Pass-2 Self-Audit Checklist attested.

Cross-refs. → synthesis-summary.md · → political-style-guide.md §Narrative Integrity · → analytical-supplementary-methodology.md §media-framing · → electoral-domain-methodology.md §Part 4 · → forward-indicators.md · → scenario-forecast.md.

comparative-international

Purpose. Cross-jurisdictional comparison: EP vs US Congress / UK Parliament / G7 legislatures / Council of Europe.

Required sources. External authoritative: US Congress.gov, UK Hansard, G7 communiqués, CoE Parliamentary Assembly. Cite with Admiralty grades.

Construction steps. (1) Identify comparable mechanism/procedure. (2) Side-by-side table: institutional design + typical outcome + duration + transparency. (3) Takeaways: what EP could learn + what's EU-unique.

Quality signals. ≥270 lines. ≥3 jurisdictions. Side-by-side matrix ≥15 rows. Takeaways with named recommendations.

Cross-refs. → implementation-feasibility.md · → historical-parallels.md.

cross-reference-map

Purpose. Inter-artifact citation map for this run. Documents citation edges and identifies orphan / over-cited artifacts.

Required sources. All artifacts in this run.

Construction steps. (1) Parse every artifact for [[link]] / ] references. (2) Build citation-edge list. (3) Mermaid graph. (4) Identify orphans and over-cites. (5) Flag evidence-terminal vs synthesis-terminal nodes.

Quality signals. ≥260 lines. Citation graph ≥30 edges. Orphan / over-cite flags. Evidence-terminal nodes identified.

Cross-refs. → structural-metadata-methodology.md · → analysis-index.md.

data-download-manifest

Purpose. Stage-A inventory of every EP MCP call + raw artifact paths + Admiralty grades. Provenance-first companion to mcp-reliability-audit.md.

Required sources. All MCP calls made during the run.

Construction steps. (1) One row per MCP call: tool, params, timestamp, artifact path, status, Admiralty grade. (2) Summary statistics. (3) Raw-artifact inventory. (4) Gaps section.

Quality signals. ≥250 lines. ≥20 MCP-call rows. Freshness stats. Gaps explicitly enumerated.

Cross-refs. → mcp-reliability-audit.md · → structural-metadata-methodology.md.

voter-segmentation

Purpose. EU electorate segmentation across 27 MS + 14 EUPM languages. Segments by Europhile/Euroskeptic, North/South/V4/Baltic/Nordic blocs, urban/rural, age cohort, education.

Required sources. Eurobarometer standard + special reports, World Bank demographics (non-economic), IMF WEO (economic context where relevant), EP 2024 election turnout by MS.

Construction steps. (1) Enumerate 6–10 segments. (2) Per segment: size, geographic density, issue salience. (3) Map segments to political-group support. (4) Identify swing segments for 2029 EP election. (5) 14-language accessibility note.

Quality signals. ≥300 lines. ≥6 segments fully profiled. Swing-segment register. 14-language accessibility note.

Mandatory SATs. Stakeholder Mapping (§4.2 supporting, primary — segment profiling is a form of citizen-group stakeholder mapping) · Key Assumptions Check (§4.1 technique 2, supporting — validate that segment stability assumptions hold for the current cycle; see voter-segmentation-methodology.md for Eurobarometer fallback rules). See osint-tradecraft-standards.md §4.

Cross-refs. → electoral-domain-methodology.md · → stakeholder-map.md.

data-availability-assessment

Purpose. Stage-A triage artifact required for every article-generating run. Records the per-source availability status of every EP MCP endpoint, IMF SDMX API, DOCEO XML (voting records), and World Bank data source for the current run. Sets manifest.dataMode which drives degraded-floor selection in Stage C. Produced before any Stage-B analysis artifact.

Required sources. All EP MCP tool calls attempted in Stage A: get_voting_records, get_latest_votes, get_adopted_texts, get_procedures_feed, get_procedures, get_current_meps, get_plenary_sessions, get_committee_documents_feed. IMF probe via scripts/imf-mcp-probe.sh. World Bank MCP calls if used.

Construction steps. (1) Call each EP MCP tool with a date-bounded probe query. Record outcome (full / partial / empty / error). (2) Probe IMF SDMX endpoint — record status (live / F6 / timeout). Check data/cache/imf/ for prior-run cache. (3) Check get_voting_records and get_latest_votes for RCV data freshness — compute lag in weeks. (4) Check get_procedures_feed for STALENESS_WARNING. (5) Select dataMode per the decision table. (6) Write manifest.dataMode field. (7) Produce Mermaid source-availability diagram.

Quality signals. ≥80 lines. Per-source triage table with ≥8 rows. dataMode field explicitly declared. manifest.json update noted. Mermaid flowchart present.

Cross-refs. → mcp-reliability-audit.md · → reference-quality-thresholds.json §degradedFloorFactors · → intelligence/voting-patterns.degraded.md (when degraded-voting) · → intelligence/economic-context.fallback.md (when degraded-imf).

procedures-proxy

Purpose. Companion artifact to mcp-reliability-audit.md produced whenever get_procedures_feed returns a STALENESS_WARNING (1972–1987 tail data). Documents the mitigation chain applied (get_procedures paginated list + get_adopted_texts proxy) so individual Stage-B artifacts do not need to re-document it.

Required sources. get_procedures_feed (outcome: STALENESS_WARNING captured). get_procedures (paginated list — current data). get_adopted_texts (outcome-proxy for procedure completions). Optionally get_meeting_decisions (session-level procedure decisions).

Construction steps. (1) Record the get_procedures_feed STALENESS_WARNING (era, item count, date). (2) Confirm this is the known EP upstream pattern (not a new defect). (3) Call get_procedures paginated list — record current procedures in scope. (4) Call get_adopted_texts — record adopted-text outcomes matching procedures. (5) Produce Mermaid mitigation-chain diagram. (6) Record Admiralty grades for fallback sources (A2 for both). (7) Summarise impact on each Stage-B artifact.

Quality signals. ≥60 lines. Staleness incident record table present. Mermaid flowchart of mitigation chain present. ≥2 fallback source rows. Impact table covering ≥3 Stage-B artifacts.

Cross-refs. → mcp-reliability-audit.md §Data-source bridge · → data-availability-assessment.md §5 Procedures-Feed Staleness Note · → .github/prompts/07-mcp-reference.md §11.

🧭 Folder Variants

Two folder-name variants appear in historical runs and are treated as equivalent by the validator:

Canonical folder Historical variant Treatment
risk-scoring/ risk/ Both accepted. New runs must write to risk-scoring/. The validator resolves risk/ reads for backward compatibility.
intelligence/ existing/ (for long-form) Both accepted for deep-analysis.md, session-baseline.md, and mirror artifacts. New runs may use either; intelligence/ is preferred.

When authoring a run from scratch, always use the canonical folder names; when reading prior runs for cross-run-diff or historical-baseline, resolve both variants.

🔗 Related Documents

Document Control:

  • Path: /analysis/methodologies/per-artifact-methodologies.md
  • Classification: Public
  • Version: 1.5 — v1.5 (2026-05-14) adds **Mandatory SATs.** lines to every artifact section (primary + supporting SATs now grep-able as ^\*\*Mandatory SATs\.\*\*); adds cross-references to new methodology pages source-triangulation.md, confidence-calibration.md, and voter-segmentation-methodology.md; updates the Cross-Cutting Tradecraft Reference section to reflect the per-artifact SAT contract. v1.4 (2026-04-25, late) adds the ### threat-analysis section to match the new threat-assessment/threat-analysis.md row in artifact-catalog.md (group now 5/5 reconciled). v1.3 (2026-04-25) refreshes dates and tightens cross-references for the v3.2 methodology release; no ### artifact section was added or removed. v1.2 (2026-04-23) added the extended/ folder with 12 optional deep-intelligence artifacts (executive-brief, devils-advocate-analysis, historical-parallels, coalition-mathematics, forward-indicators, intelligence-assessment, implementation-feasibility, media-framing-analysis, comparative-international, cross-reference-map, data-download-manifest, voter-segmentation) from the riksdagsmonitor port. v1.1 added voting-patterns, workflow-audit, cross-session-intelligence, deep-analysis, session-baseline and documented the existing/ legacy folder + risk/ folder variant. v1.0 was the initial per-artifact construction rules extracted from Run 184 reference benchmark.
  • Next Review: 2026-07-31