collect-evidence
Skill0 starsChanged 4 months agoMITView on GitHub
Search real sources for material bearing on the question.
4 other projects use a version of this.
Adopted by
Most recently taken up by
- Greg-Klein/implementation-factorytheir own version, first seen Oct 11, 2026
- Greg-Klein/implementation-harnesstheir own version, first seen Oct 9, 2026
- yezhwi/superpowers-engineering-harnesstheir own version, first seen Sep 29, 2026
- Omerfaruk-aydn/Agent_OStheir own version, first seen Jul 20, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- yezhwi/superpowers-engineering-harness★ 3same name
- Greg-Klein/implementation-factory★ 1same name
- Greg-Klein/implementation-harness★ 1same name
- Omerfaruk-aydn/Agent_OS★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- analyze-competitionSkill
- analyze-marketSkill
- cluster-painsSkill
- compose-reportSkill
- contradiction-analysisSkill
- decisionSkill
- discover-opportunitiesSkill
- harvest-leadsSkill
- interview-planSkill
- research-briefSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 1 is distinctive, and 0 are very specific to this project.
0
Common
Most of the 4 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 30
Keep an item when it is first-hand about the problem space — someone describing their own workflow, their own frustration, what they tried, …
“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Collect evidence
Real material has already been fetched from external sources. Your job is to decide which of it bears on the question, and to quote the passage that carries the point.
You are selecting, not writing. Every excerpt you keep is checked against the text it came from; anything not literally present there is rejected and the whole reply fails. You cannot introduce a source, a URL, a statistic or a quotation that is not below.
The question
${question}What the sources returned
${candidates}What to keep
Keep an item when it is first-hand about the problem space — someone describing their own workflow, their own frustration, what they tried, what it cost them. That is the material the rest of the pipeline is built to reason over.
Drop:
- marketing copy, launch announcements, and anything written to sell something;
- generic commentary that could have been written without doing the work;
- items that mention the topic but say nothing about how anyone actually works;
- duplicates and cross-posts — keep the fullest telling, not both.
Relevance beats volume. Four strong items is a better result than twenty weak ones, and an empty selection is legitimate when nothing fetched genuinely bears on the question. Padding the list is the failure mode here, not missing something.
Choosing the excerpt
Copy the passage exactly as it appears in that candidate's text. Reflowing
across lines is fine; changing words, tidying, cutting mid-sentence or joining two
separate passages is not.
Pick the sentence or short paragraph a reader could see on its own and understand what the problem is. A quote that needs the surrounding thread to make sense is a poor excerpt — prefer a different item over a quote that cannot stand up alone.
candidate_id must match an external_id above exactly.
Grading each one
evidence_kind— what shape of observation this is.evidence_level— how well grounded. One person's account isanecdotal; several independent people saying the same thing iscorroborated; a figure someone measured ismeasured. Most of what you see here will beanecdotal, and saying so costs nothing.confidence— how sure you are it bears on the question, not how strongly the author felt.relevance— one line on what it shows. Not a summary of the item; a statement of why it matters to the question asked.
What would make this stage worthless
Selecting an item because it is about the right topic rather than because it shows
someone struggling. Grading everything corroborated because several items mention
the same tool. Quoting the most quotable line rather than the most informative one.
Keeping something because the list would otherwise look thin.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

