fix: per-context SCTR output files + fixed Q3.2 constraints sub-level (#531, #532)#610
Merged
Merged
Conversation
…el (LLM-Coding#531, LLM-Coding#532) LLM-Coding#531 — Phase 1 wrote fixed filenames (QUESTION_TREE.adoc, OPEN_QUESTIONS.adoc), so sequential bounded-context runs silently overwrote each other while the skill itself documents "one bounded context at a time". Phase 1 outputs are now suffixed with the kebab-cased context name, Phase 2 reads the context-specific tree, and ADRs get a context prefix (adrs/[context-name]-adr-NNN-*.adoc) so two contexts cannot produce colliding decision-record files. All references to the old fixed names updated (SKILL.md, both prompts, output-schema reference, website doc EN+DE). LLM-Coding#532 — the adaptive, code-density-driven decomposition under Q3.2 (Architecture Constraints) walked past constraints that do not live in dense code: process constraints (git workflow), convention constraints (naming, mandated libraries), runtime constraints. Q3.2 now carries a fixed third level mirroring arc42 Chapter 2's constraint kinds — Q3.2.1 technical, Q3.2.2 organizational/process, Q3.2.3 conventional — and the prompt explicitly admits CLAUDE.md/AGENTS.md, CONTRIBUTING, CI workflows, and linter configs as evidence sources for this branch. This closes the untraceable-claim gap where Phase 2 picked such constraints up from the repo without any tree leaf to cite. Skill version 0.2 -> 0.3, plugin 0.3.0 -> 0.4.0. Fixes LLM-Coding#531, fixes LLM-Coding#532 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (2)
WalkthroughDie Änderungen bringen Phase‑1/Phase‑2 Prompts, Skill‑Metadaten und Referenzdokumentation auf kontextspezifische Ausgabe‑Dateinamen und fügen unter Q3.2 eine feste dritte Ebene (Q3.2.1–Q3.2.3) plus ausdrücklich erlaubte Nicht‑Code‑Evidenzquellen hinzu. ChangesSocratic Code-Theory Recovery v0.3: Multi-Bounded-Context-Workflow
Estimated code review effort🎯 2 (Simple) | ⏱️ ~12 minutes Possibly related PRs
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (2)
plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-1-question-tree.md (1)
51-123:⚠️ Potential issue | 🟠 Major | ⚡ Quick winEinheitliches Evidenzformat erzwingen. In beiden Phase-1-Prompt-Kopien erlauben Schritt 3/4 noch
file::function, während der spätere Sanity-Check nurfile:lineprüft. Bitte nurfile:linezulassen, damit[ANSWERED]-Leaves in beiden Prompts verlässlich prüfbar bleiben.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-1-question-tree.md` around lines 51 - 123, The prompt currently allows evidence in two formats (`file::function` and `file:line`) in steps 3/4 but the Sanity-check only validates `file:line`; update both Phase‑1 prompt copies to accept and require only the `file:line` evidence format (remove any mention of `file::function` from the decomposition and leaf classification rules and the examples), and also adjust the Sanity-check instructions to assert/verify only `file:line` citations for `[ANSWERED]` leaves (affecting text around "For each leaf, classify it" and the "What to do after the prompt completes" sanity checks and the references to QUESTION_TREE-[context-name].adoc / OPEN_QUESTIONS-[context-name].adoc).plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.md (1)
72-78:⚠️ Potential issue | 🟠 Major | ⚡ Quick winGemeinsame Traceability-Regel für beide Phase-2-Prompts anpassen.
plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.mdundskill/socratic-code-theory-recovery/prompts/phase-2-synthesize.mdbehandeln[ANSWERED]-Leaves noch als reine „code evidence“. Für Q3.2 müssen beide Prompts Nicht-Code-Belege alsfile:linezulassen, sonst gehen gültige Architektur-Constraints ausCLAUDE.md,AGENTS.md,CONTRIBUTING.mdoder CI-/Lint-Konfigurationen verloren.🔧 Vorschlag
- A claim backed by an [ANSWERED] leaf cites the code evidence from that - leaf — the reference to the code, the only durable, canonical artifact: + A claim backed by an [ANSWERED] leaf cites the evidence from that leaf + as `file:line` — code and non-code sources are both valid:Basierend auf dem PR-Ziel
#532: Nicht-Code-Evidenz muss in Phase 2 ausdrücklich erlaubt bleiben.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.md` around lines 72 - 78, Die Regel in both prompts currently treats [ANSWERED] leaves as only "code evidence"; update the text in plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.md and skill/socratic-code-theory-recovery/prompts/phase-2-synthesize.md to allow non-code evidence by permitting Evidence lines that reference non-source files in the form file:line (e.g., CLAUDE.md:12, AGENTS.md:5, CONTRIBUTING.md:3, or CI/LINT config paths), while still requiring the Evidence line be copied verbatim from the leaf and preserving the [ANSWERED] semantics for Q3.2; keep the instruction to not invent/shorten paths and add an explicit example or phrase that non-code artifacts (docs/configs/MD files) are valid file:line evidence.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@plugins/semantic-anchors/skills/socratic-code-theory-recovery/SKILL.md`:
- Line 27: Aktualisiere beide SKILL-Kopien: ersetze den aktuellen Phase-2-Guard,
der auf die Ausnahme "unless the user explicitly asks" basiert, durch eine
Prüfung der OPEN-Blätter in der Datei OPEN_QUESTIONS-<context-name>.adoc und
lasse Phase 2 nur starten, wenn alle `[OPEN]`-Leaves entweder beantwortet sind
oder explizit mit `(deferred)` markiert wurden; passe die Nutzerhinweise (die
Phase‑1‑Abschlussmeldung) in denselben Dateien an, damit sie eindeutig
mitteilen, wo die zwei Ausgabedateien liegen und dass Phase 2 erst nach
beantworteten/`(deferred)`-markierten OPEN-Blättern ausgeführt wird.
---
Outside diff comments:
In
`@plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-1-question-tree.md`:
- Around line 51-123: The prompt currently allows evidence in two formats
(`file::function` and `file:line`) in steps 3/4 but the Sanity-check only
validates `file:line`; update both Phase‑1 prompt copies to accept and require
only the `file:line` evidence format (remove any mention of `file::function`
from the decomposition and leaf classification rules and the examples), and also
adjust the Sanity-check instructions to assert/verify only `file:line` citations
for `[ANSWERED]` leaves (affecting text around "For each leaf, classify it" and
the "What to do after the prompt completes" sanity checks and the references to
QUESTION_TREE-[context-name].adoc / OPEN_QUESTIONS-[context-name].adoc).
In
`@plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.md`:
- Around line 72-78: Die Regel in both prompts currently treats [ANSWERED]
leaves as only "code evidence"; update the text in
plugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.md
and skill/socratic-code-theory-recovery/prompts/phase-2-synthesize.md to allow
non-code evidence by permitting Evidence lines that reference non-source files
in the form file:line (e.g., CLAUDE.md:12, AGENTS.md:5, CONTRIBUTING.md:3, or
CI/LINT config paths), while still requiring the Evidence line be copied
verbatim from the leaf and preserving the [ANSWERED] semantics for Q3.2; keep
the instruction to not invent/shorten paths and add an explicit example or
phrase that non-code artifacts (docs/configs/MD files) are valid file:line
evidence.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yml
Review profile: CHILL
Plan: Pro
Run ID: 6e8b9a41-4635-4d4e-ae05-97c03503153a
📒 Files selected for processing (14)
docs/changelog.adocdocs/socratic-recovery-skill.adocdocs/socratic-recovery-skill.de.adocplugins/semantic-anchors/plugin.jsonplugins/semantic-anchors/skills/socratic-code-theory-recovery/SKILL.mdplugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-1-question-tree.mdplugins/semantic-anchors/skills/socratic-code-theory-recovery/prompts/phase-2-synthesize.mdplugins/semantic-anchors/skills/socratic-code-theory-recovery/references/arc42.mdplugins/semantic-anchors/skills/socratic-code-theory-recovery/references/output-schema.mdskill/socratic-code-theory-recovery/SKILL.mdskill/socratic-code-theory-recovery/prompts/phase-1-question-tree.mdskill/socratic-code-theory-recovery/prompts/phase-2-synthesize.mdskill/socratic-code-theory-recovery/references/arc42.mdskill/socratic-code-theory-recovery/references/output-schema.md
Step 4 of the invocation flow allowed Phase 2 "unless the user explicitly asks", contradicting the Between-Phases rule and the Phase 2 prompt header, which both require every [OPEN] leaf to carry a team answer or an explicit (deferred) marker. Phase 2 is now gated on that file check — even an explicit user request first verifies OPEN_QUESTIONS-<context-name>.adoc and lists the unanswered leaves instead of proceeding. Found by CodeRabbit review. Refs LLM-Coding#531, LLM-Coding#532 Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Fixes the two Socratic Code-Theory Recovery bugs found during the Mistral Vibe brownfield runs.
#531 — Phase 1 output collision
Phase 1 wrote two fixed filenames, so the documented "one bounded context at a time" workflow silently destroyed earlier runs. Now:
QUESTION_TREE-[context-name].adoc/OPEN_QUESTIONS-[context-name].adoc(kebab-cased context name, consistent with Phase 2's existingprd-[context-name].adocscheme)docs/specs/adrs/[context-name]-adr-NNN-*.adoc— closes the secondary collision in the sharedadrs/folderreferences/output-schema.md, and the website doc page (EN + DE). Grep-verified: no un-parameterizedQUESTION_TREE.adoc/OPEN_QUESTIONS.adocreference remains.#532 — Q3.2 constraints under-decomposition
The code-density-driven decomposition walked past constraints that do not live in dense code (git workflow, naming conventions, mandated libraries, runtime choices) — Phase 2 then picked them up from the repo with no tree leaf to cite, producing exactly the untraceable claims the two-phase design exists to prevent. Now Q3.2 carries a fixed third level, mirroring arc42 Chapter 2's constraint kinds:
For this branch specifically, the prompt admits non-code evidence sources —
CLAUDE.md/AGENTS.md, CONTRIBUTING, CI workflows, linter/formatter configs — cited file:line like any other evidence.references/arc42.mdcarries the same note.Versioning
Skill
0.2→0.3, plugin0.3.0→0.4.0(same PR, per convention).Note on the contract
The Socratic Code Theory Recovery contract text (CLAUDE.md / contracts.json) states "the second level of the tree is FIXED" — the fixed Q3.2.x third level is an extension of that idea. If the contract should mirror it, that is a separate edit to contracts.json (and your CLAUDE.md); happy to follow up.
Verification
node scripts/render-docs.js: website doc pages (EN + DE) render cleanFixes #531, fixes #532
🤖 Generated with Claude Code
Summary by CodeRabbit
Documentation
QUESTION_TREE-<context-name>.adoc,OPEN_QUESTIONS-<context-name>.adoc) — verhindert Überschreiben zwischen Läufen; Phase‑1 stoppt bis Teamantwort/(deferred)-Marker.Chores