Skip to content

Commit f9bd723

Browse files
justfinethankuJim Powers
justfinethanku
authored and
Jim Powers
committed
[skills] Add weekly signal diff skill pack (NateBJones-Projects#178)
* [skills] Add weekly signal diff skill pack * [skills] Fix markdownlint numbering in weekly signal diff
1 parent 297bec8 commit f9bd723

5 files changed

Lines changed: 401 additions & 0 deletions

File tree

‎skills/weekly-signal-diff/README.md‎

Lines changed: 107 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,107 @@
1+
# Weekly Signal Diff
2+
3+
> Standalone skill pack for turning a week's worth of market noise into a
4+
> personalized structural diff.
5+
6+
## What It Does
7+
8+
This skill runs a weekly scan across a suggested universe of categories and
9+
companies, then reweights the analysis using what Open Brain already knows
10+
about the user. It produces a `diff, not digest`: what changed, why it
11+
matters, and what to watch next.
12+
13+
The default starter universe is AI-first because that is where the original
14+
use case came from, but the logic is universal. You can swap the categories and
15+
companies for any fast-moving market and keep the same structural-diff process.
16+
17+
## Supported Clients
18+
19+
- Claude Code
20+
- Codex
21+
- Cursor
22+
- Other AI clients that support reusable prompt packs, rules, or custom
23+
instructions
24+
25+
## Prerequisites
26+
27+
- Working Open Brain setup if you want memory search and capture
28+
([guide](../../docs/01-getting-started.md))
29+
- AI client that supports reusable skills, rules, or custom instructions
30+
- One of:
31+
- live web access in the client
32+
- a user-provided weekly source set
33+
- Optional upgrade for automated live-search runs: OpenRouter access to the
34+
Perplexity Sonar family
35+
([OpenRouter model page](https://openrouter.ai/perplexity/sonar/api))
36+
37+
## Installation
38+
39+
1. Copy [`SKILL.md`](./SKILL.md) into your client's reusable-instructions
40+
location.
41+
2. Keep the [`references/`](./references/) folder alongside it if you want the
42+
starter universe and live-search notes available to the client.
43+
3. Restart or reload the client so it picks up the skill.
44+
4. Test it with a prompt like:
45+
`Run my weekly signal diff on AI infrastructure and tell me what changed this week that matters for a solo builder.`
46+
5. Optional: wire it into your weekly automation. If OpenRouter is available,
47+
use the Perplexity Sonar family for the retrieval pass and keep the final
48+
digest structure consistent every week.
49+
50+
For Claude Code, a common install path is:
51+
52+
```bash
53+
mkdir -p ~/.claude/skills/weekly-signal-diff/references
54+
cp skills/weekly-signal-diff/SKILL.md ~/.claude/skills/weekly-signal-diff/SKILL.md
55+
cp -R skills/weekly-signal-diff/references ~/.claude/skills/weekly-signal-diff/references
56+
```
57+
58+
If your client does not support native skill folders, paste the contents of
59+
[`SKILL.md`](./SKILL.md) into that client's reusable prompt or project-rules
60+
feature and keep the reference files nearby.
61+
62+
## Trigger Conditions
63+
64+
- "Run my weekly signal diff"
65+
- "What changed this week that matters to me?"
66+
- "Track AI this week"
67+
- "Turn this week's news into structural shifts"
68+
- "Give me the signal, not the headlines"
69+
- Weekly automated digests or review rituals
70+
71+
## Expected Outcome
72+
73+
When installed and invoked correctly, the skill should produce:
74+
75+
- a coverage note explaining what was scanned
76+
- 3-7 structural shifts instead of a long news list
77+
- user-specific implications pulled from Open Brain memory
78+
- a watchlist for next week
79+
- optional capture of the weekly digest back into Open Brain
80+
81+
## Troubleshooting
82+
83+
**Issue: The output reads like a news summary**
84+
Solution: Keep the structural questions intact. The skill should filter for
85+
constraint shifts, leverage shifts, broken assumptions, and exposed
86+
dependencies.
87+
88+
**Issue: The final diff feels generic**
89+
Solution: Check that the client actually searched Open Brain first. This skill
90+
gets sharper when it can pull active projects, recurring interests, and prior
91+
digests before ranking the week's news.
92+
93+
**Issue: The scan fixates on the default 30-company list**
94+
Solution: Treat the starter universe as a bootstrap layer. Replace or re-rank
95+
the suggested companies and categories using the user's actual focus areas.
96+
97+
**Issue: The live-search results are shallow or stale**
98+
Solution: If OpenRouter is available, upgrade the retrieval pass to a
99+
Perplexity Sonar search model and constrain domains or freshness when needed.
100+
See [references/live-search-upgrade.md](./references/live-search-upgrade.md).
101+
102+
## Notes for Other Clients
103+
104+
This skill is portable because the logic is procedural. Any client that can
105+
load reusable instructions and access either Open Brain or a weekly source set
106+
can run it. If the client has no live search, feed it a source packet and ask
107+
for the same structural-diff output.

‎skills/weekly-signal-diff/SKILL.md‎

Lines changed: 158 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,158 @@
1+
---
2+
name: weekly-signal-diff
3+
description: |
4+
Use when the user wants a weekly structural diff on AI, software, or another
5+
fast-moving market. Starts from 10 suggested categories and 30 suggested
6+
companies when no watchlist exists, then adapts the scan using Open Brain
7+
memory, current priorities, and prior digests. Best for prompts like "run my
8+
weekly signal diff", "what changed this week that matters to me", "track this
9+
market", or "turn this week's news into structural shifts". Optional live
10+
search upgrade: if OpenRouter access is available, prefer the Perplexity
11+
Sonar family for fresh web-grounded retrieval with citations.
12+
author: Jonathan Edwards
13+
version: 1.0.0
14+
---
15+
16+
# Weekly Signal Diff
17+
18+
## Problem
19+
20+
A wall of news does not tell the user what structurally changed. Most weekly
21+
roundups over-index on headlines, underweight economics and dependency shifts,
22+
and ignore what the user actually cares about. This skill turns a noisy week
23+
into a small set of structural changes, weighted by Open Brain memory.
24+
25+
## When to Use
26+
27+
- Weekly market review or Sunday/Friday ritual
28+
- "Run my weekly signal diff"
29+
- "What changed this week that matters to me?"
30+
- "Track this market and tell me the structural shifts"
31+
- "Turn this pile of news into a decision-grade diff"
32+
- Ongoing automation that writes a weekly digest back to Open Brain
33+
34+
## Required Context
35+
36+
Gather as much as the environment allows:
37+
38+
- the user's active projects, bets, and recurring interests
39+
- prior weekly digests or adjacent summaries stored in Open Brain
40+
- the desired freshness window (default: last 7 days)
41+
- any preferred outlets, banned sources, or explicit watchlist entities
42+
43+
If the user has not provided categories or companies, read
44+
[references/starter-universe.md](references/starter-universe.md) and use it as
45+
a bootstrap layer only.
46+
47+
If live web access is available and the user wants current coverage, read
48+
[references/live-search-upgrade.md](references/live-search-upgrade.md) and use
49+
the strongest search mode the environment supports.
50+
51+
## Process
52+
53+
1. Establish the frame.
54+
- Confirm the topic space, freshness window, and whether the goal is
55+
personal awareness, operator strategy, investor tracking, or content prep.
56+
- If the user says nothing, default to a 7-day operator-style review.
57+
58+
1. Pull Open Brain context first.
59+
- Search for active projects, current priorities, recurring entities, recent
60+
captures, and the last 2-4 weekly digests.
61+
- Tool names vary by client. Use the available Open Brain search, list, and
62+
capture tools in the environment rather than assuming fixed names.
63+
- Extract a short relevance profile: what the user is building, what they
64+
keep revisiting, what they are worried about, and what they are trying to
65+
learn.
66+
67+
1. Build the watchlist.
68+
- Start from the suggested 10-category / 30-company starter universe if the
69+
user has not defined a watchlist.
70+
- Treat the starter list as a scaffold, not a contract.
71+
- Re-rank or replace items using Open Brain context:
72+
- promote companies, categories, or themes the user mentions often
73+
- demote low-signal items
74+
- add personal-priority entities even if they are outside the starter set
75+
- Preserve some baseline discovery. Personalization should shape the scan,
76+
not collapse it into only known favorites.
77+
78+
1. Gather the week's evidence.
79+
- Prefer fresh, source-backed information with links or citations.
80+
- If live search is available, perform a broad sweep first, then targeted
81+
follow-ups on the top candidate shifts.
82+
- If live search is not available, work from the user's provided sources and
83+
say that the diff is source-bounded.
84+
- Ignore pure announcement theater unless it changes economics,
85+
distribution, regulation, dependency, geography, or buyer behavior.
86+
87+
1. Ask the structural questions on every candidate signal.
88+
- What constraint shifted?
89+
- Who gained or lost leverage?
90+
- What got cheaper, harder, faster, or more defensible?
91+
- What dependency got exposed?
92+
- What business model or pricing assumption weakened?
93+
- What changed in regulation, geography, or distribution?
94+
- Why does this matter for the user's actual projects, workflows, or market
95+
view?
96+
97+
1. Score before writing.
98+
- Keep only the few signals that represent real change.
99+
- A good weekly diff usually has 3-7 structural shifts.
100+
- Merge duplicates, drop weak stories, and explicitly label speculation as
101+
speculation.
102+
103+
1. Produce the weekly diff.
104+
105+
Use this default structure:
106+
107+
- `Coverage note` — what was scanned, how it was personalized, and the date
108+
window
109+
- `Structural shifts` — 3-7 items, each with:
110+
- what changed
111+
- why it matters in general
112+
- why it matters to this user
113+
- supporting evidence or citations
114+
- `What changed from last week` — new, rising, fading, or resolved themes
115+
- `Watch next` — entities, constraints, or questions to monitor
116+
- `Actions` — optional follow-ups, only if the evidence supports them
117+
118+
1. Capture the durable output.
119+
- Save the final digest back into Open Brain when capture tools are
120+
available.
121+
- Prefer one durable weekly summary plus separate captures only for truly
122+
important follow-up items.
123+
- Include provenance: week ending date, topic scope, and major entities
124+
covered.
125+
126+
## Output
127+
128+
When this skill works correctly, the user gets:
129+
130+
- a concise weekly structural diff instead of a headline roundup
131+
- a clear explanation of why the shifts matter to them specifically
132+
- citations or source links when live search is available
133+
- a durable weekly digest saved back into Open Brain for future comparison
134+
135+
## Guardrails
136+
137+
- The goal is `diff, not digest`.
138+
- Do not force all 30 suggested companies into the final output. They are there
139+
to prevent blank-page syndrome, not to create fake coverage.
140+
- Do not mistake product launches, benchmark screenshots, or funding headlines
141+
for structural change unless they move a real constraint.
142+
- Keep general market analysis separate from personalized implications.
143+
- If evidence is thin, say the week was thin.
144+
- If the environment lacks live search, be explicit about the freshness
145+
limitation.
146+
- If the user's interests are unclear, use the starter universe and explain
147+
that it is a bootstrap pass.
148+
149+
## Notes for Other Clients
150+
151+
- This skill is portable across Claude Code, Codex, Cursor, and similar clients
152+
because the core behavior is procedural.
153+
- Adapt Open Brain tool names to the local environment.
154+
- For scheduled runs, pair the skill with the user's automation system and keep
155+
the same structure every week so diffs stay comparable.
156+
- If OpenRouter is available, prefer a Perplexity Sonar web-search model for
157+
the retrieval pass, then use the local AI client or model to do the actual
158+
synthesis if that split is more ergonomic.

‎skills/weekly-signal-diff/metadata.json‎

Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,27 @@
1+
{
2+
"name": "Weekly Signal Diff",
3+
"description": "Standalone skill pack for turning a week's worth of market or AI news into a personalized structural diff using Open Brain memory, a suggested starter universe, and optional live web search.",
4+
"category": "skills",
5+
"author": {
6+
"name": "Jonathan Edwards",
7+
"github": "justfinethanku"
8+
},
9+
"version": "1.0.0",
10+
"requires": {
11+
"open_brain": true,
12+
"services": ["Optional: OpenRouter (Perplexity Sonar family) for live search"],
13+
"tools": ["Claude Code or similar AI coding tool with reusable skills/system prompts"]
14+
},
15+
"tags": [
16+
"weekly-review",
17+
"signal-diff",
18+
"market-intelligence",
19+
"news",
20+
"openrouter",
21+
"research"
22+
],
23+
"difficulty": "intermediate",
24+
"estimated_time": "15 minutes",
25+
"created": "2026-04-13",
26+
"updated": "2026-04-13"
27+
}

‎skills/weekly-signal-diff/references/live-search-upgrade.md‎

Lines changed: 57 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,57 @@
1+
# Live Search Upgrade
2+
3+
Use this file only when the environment supports current web retrieval and the
4+
user wants fresh, source-backed coverage.
5+
6+
## Preferred Upgrade Path
7+
8+
For OB1 users with OpenRouter access, prefer the Perplexity Sonar family for
9+
the retrieval pass. Start with the strongest Sonar search tier available in the
10+
user's account. Use the plain Sonar tier as the lowest-cost fallback when
11+
budget matters more than depth.
12+
13+
Useful references:
14+
15+
- [OpenRouter: Perplexity Sonar](https://openrouter.ai/perplexity/sonar/api)
16+
- [Perplexity Sonar docs](https://docs.perplexity.ai/docs/sonar/models/sonar)
17+
- [Perplexity search filters](https://docs.perplexity.ai/docs/grounded-llm/chat-completions/filters/academic-filter)
18+
19+
## Retrieval Pattern
20+
21+
Run the search in two passes:
22+
23+
1. Broad sweep
24+
- scan the last 7 days across the suggested categories and high-priority
25+
entities
26+
- return only source-backed developments with links or citations
27+
- shortlist the stories that look like structural change
28+
29+
2. Targeted follow-up
30+
- deepen only the top 3-7 candidate shifts
31+
- tighten recency, domain filters, or entity filters when needed
32+
- pull enough evidence to explain both the general impact and the personal
33+
relevance
34+
35+
## What to Ask the Search Layer For
36+
37+
Prefer prompts or search instructions that request:
38+
39+
- a 7-day freshness window unless the user says otherwise
40+
- cited links or explicit source URLs
41+
- domain filters when the user trusts a specific set of outlets
42+
- one-paragraph explanations of why each result matters
43+
- rejection of results that are just launch noise or funding theater
44+
45+
## Automation Notes
46+
47+
For scheduled runs:
48+
49+
- keep the retrieval and synthesis structure consistent every week
50+
- store the final digest back in Open Brain so next week's run has something to
51+
diff against
52+
- track the week-ending date in the saved summary
53+
- if cost matters, use a cheaper broad sweep and spend extra search depth only
54+
on the top candidate shifts
55+
56+
The search layer finds the evidence. Open Brain decides what matters to this
57+
user. Keep those roles separate.

0 commit comments

Comments
 (0)