facts-ext
Skill0 starsChanged 4 months agoMITView on GitHub
Research a topic using web, filesystem, and inference with parallel sub-agents. Usage: /facts-ext {topic} [hints...]
In this repository21 other files it gives its agents.
- add-to-claudeSkill
- add-to-rulesSkill
- audit-claudeSkill
- audit-rulesSkill
- autopilotSkill
- better-planSkill
- design-autoSkill
- designSkill
- execute-autoSkill
- executeSkill
- factsSkill
- questionsSkill
- research-autoSkill
- researchSkill
- review-autoSkill
- structure-autoSkill
- structureSkill
- wiki-captureSkill
- wiki-ingestSkill
- wiki-lintSkill
- wiki-querySkill
The file
We split this skill into the rules it gives an agent
11 rules. About 0 are common, 9 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 20
Use the Agent tool with subagent_type: "ext-fact-finder".
- Distinctive: line 28
After the root agent completes, seed both lists from its response:
- Distinctive: line 29
Add its COVERED items to covered
- Distinctive: line 30
Add its SUGGESTED items to queue
- Distinctive: line 37
If the filtered queue is empty, proceed to Phase 3.
- Distinctive: line 45
After all agents in this round complete, for each agent:
- Distinctive: line 47
Add its SUGGESTED items to queue (replacing the old queue)
- Distinctive: line 52
After the loop exits:
- Very specific: line 57
Use # {Topic} — Research Facts as the top heading
- Distinctive: line 61
No frontmatter, no timestamps, no filler
- Distinctive: line 62
Delete all fragment files
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Parse the user's input:
- First argument is the topic (e.g.,
Postgres advisory lock semantics,S3 pre-signed URL expiry rules,OAuth device-flow constraints) - Any additional arguments are hints — URLs, file paths, or narrowing keywords the agent should start with
Phase 1 — Root research
Spawn an ext-fact-finder agent with:
topic: {topic}
sub_topic: {topic}
hints: {any hints provided, or "none"}Use the Agent tool with subagent_type: "ext-fact-finder".
Phase 2 — Recursive fan-out
Maintain two lists across rounds:
- covered — all topics explored so far (accumulated from every agent's COVERED section)
- queue — sub-topics to explore next
After the root agent completes, seed both lists from its response:
- Add its COVERED items to covered
- Add its SUGGESTED items to queue
Then loop:
- Filter the queue. For each proposed sub-topic, evaluate:
- Relevance: Is this sub-topic meaningfully related to the original topic
{topic}? Drop it if it has drifted to a tangential or generic area. - Novelty: Has this topic (or something substantially overlapping) already been covered? Drop it if so.
- Relevance: Is this sub-topic meaningfully related to the original topic
- If the filtered queue is empty, proceed to Phase 3.
- Spawn one
ext-fact-finderagent per sub-topic in the queue, all in parallel:
Use the Agent tool withtopic: {topic} sub_topic: {sub-topic name} hints: {hints from the suggestion}subagent_type: "ext-fact-finder". - After all agents in this round complete, for each agent:
- Add its COVERED items to covered
- Add its SUGGESTED items to queue (replacing the old queue)
- Go back to step 1.
Phase 3 — Merge & clean up
After the loop exits:
- Glob
docs/{topic}/facts-*.mdto find all fragment files - Read every fragment
- Combine all fragments into
docs/{topic}/facts.md:- Use
# {Topic} — Research Factsas the top heading - Organize sections logically (root agent's broad content first, then deeper sub-topic sections)
- Deduplicate facts that appear in multiple fragments
- Preserve all source citations
- No frontmatter, no timestamps, no filler
- Use
- Delete all fragment files
Tell the user the output is at docs/{topic}/facts.md.
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.

