knowledge-capture
Skill8 starsChanged 4 days agoMITView on GitHub
Use when capturing SME expertise through structured interviewing (proactive), or formalizing conversations/debug sessions/tribal knowledge into durable wiki documentation (reactive). Bottom-line-up-front articles with provenance, published to your team's wiki. NOT for editing existing wiki pages, design exploration, or casual Q&A.
9 other projects use a version of this.
Adopted by
Most recently taken up by
- tderoussel/agent-skillstheir own version, first seen Nov 1, 2026
- ravidsrk/agent-skillstheir own version, first seen Nov 1, 2026
- NoobyGains/godmodetheir own version, first seen Nov 1, 2026
- NashEQify/forgetheir own version, first seen Oct 25, 2026
- ARCASSystems/FounderOStheir own version, first seen Oct 24, 2026
- hangljoe/AI-PreSales-Skillstheir own version, first seen Sep 28, 2026
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- NoobyGains/godmode★ 120same name
- basicmachines-co/basic-memory-plugins★ 9same name
- ARCASSystems/FounderOS★ 7same name
- Empty8492/knowledge-capture★ 2same name
- NashEQify/forge★ 2same name
- hangljoe/AI-PreSales-Skills★ 1same name
- ravidsrk/agent-skills★ 1same name
- tderoussel/agent-skills★ 1same name
- stevancris/sre-ai-agent★ 0same 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 repository134 other files it gives its agents.
- AGENTS.mdAGENTS.md
- guidance/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- guidance/CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- .cursor/rules/cli-tools.mdcCursor rule
- .cursor/rules/invariants.mdcCursor rule
- .cursor/rules/javascript.mdcCursor rule
- .cursor/rules/security.mdcCursor rule
- .cursor/rules/shell.mdcCursor rule
- blast-radius-checkSkill
- brainstormingSkill
- branch-flow-gateSkill
- branch-sync-gateSkill
- codebase-reconSkill
- codeowners-drift-auditSkill
- code-review-batterySkill
- cognitive-complexity-refactoringSkill
- debateSkill
- debug-conductorSkill
- dispatching-parallel-agentsSkill
- domain-buildSkill
- evidence-adjudicatorSkill
- executing-plansSkill
- external-cli-auditSkill
- feature-developmentSkill
- field-rename-verificationSkill
- finishing-a-development-branchSkill
- git-branch-conventionsSkill
- gitlab-cliSkill
- hotfix-charterSkill
- implementation-trackerSkill
- incident-revert-reviewSkill
- infra-config-investigatorSkill
- investigation-stateSkill
- kernel-splitSkill
- llm-behavior-investigatorSkill
- llm-skill-reviewSkill
- merge-authorization-gateSkill
- micro-harsh-reviewSkill
- output-verificationSkill
- pre-commit-gateSkill
- progressive-code-review-gateSkill
- progressive-harsh-reviewSkill
- providing-code-reviewSkill
- pr-triage-gateSkill
- push-authorization-gateSkill
- receiving-code-reviewSkill
- reproduction-experiment-investigatorSkill
- requesting-code-reviewSkill
- requirements-validation-pmSkill
- requirements-validationSkill
- scope-tripwireSkill
- session-handoffSkill
- skills-hierarchy-tuningSkill
- sp-bughuntSkill
- state-consistency-investigatorSkill
- subagent-driven-developmentSkill
- systematic-debuggingSkill
- test-driven-developmentSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
12 rules. 0 are common, 10 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
10
Distinctive
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 66
Ask: What domain/topic? Who will read this?
- Distinctive: line 68
New-vs-update check: Search your team's wiki for existing pages on this topic.
- Very specific: line 69
Determine entry mode: if the conversation already contains substantive domain content AND the user's trigger was reactive -> set Source: …
- Distinctive: line 72
HARD GATE: Do not proceed without: topic, audience, intent, scope, entry mode, and new-vs-update decision.
- Distinctive: line 86
HARD GATE: Do not skip this phase in reactive mode.
- Distinctive: line 93
Reactive mode shortcut: if Phase 1.5 covered all P0 areas and >=3 P1 areas, ask: "Proceed to synthesis or do you want follow-up questions?"
- Distinctive: line 94
Run question loop: one question at a time, driven by highest-priority uncovered area.
- Distinctive: line 96
After each answer: update coverage, tag provenance, check for contradictions.
- Distinctive: line 99
HARD GATE: Do not proceed to Phase 2.5 without sufficiency gate passing OR explicit user stop.
- Distinctive: line 133
Run pre-publish checks: secret scan, duplicate detection.
- Distinctive: line 134
Run placement algorithm. Ask interviewee to confirm location.
- Distinctive: line 135
Ask for explicit publish approval.
“Common” is measured: kept by at least 6 in 10 of the 8 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
knowledge-capture
Announce at start: "I'm using the knowledge-capture skill to capture expertise and produce a wiki article."
Routing
| Signal | Wrong Skill | Right Skill |
| "interview me about..." | wiki editing | knowledge-capture |
| "turn this conversation into a wiki article" | brainstorming | knowledge-capture |
| "edit this wiki page" | knowledge-capture | wiki editing |
| "write a design doc" | knowledge-capture | brainstorming |
Collision note: expert-interviewer shares "interview me about". knowledge-capture takes precedence — it is the heavier-duty, wiki-integrated version.
Module Loading (MANDATORY)
| Module | Load BEFORE | If unavailable, use: |
modules/state-format.md | Any state file read/write | Track state in a Markdown file with sections: Topic, Phase, Source (interview/conversation), Wiki Page ID, Interview Log (append-only Q&A) |
modules/coverage-matrix.md | Starting Phase 1.5 or Phase 2 | Use areas: Context/Background, Core Concepts, How it Works, Failure Modes, Common Mistakes, Expert Tips, References. Mark each: open/partial/covered. |
modules/bluf-template.md | Starting Phase 3 (drafting) | Article format: ## Bottom Line (1-2 sentences), ## Background, ## How It Works, ## Failure Modes, ## References + Source Notes |
modules/review-rubric.md | Starting Phase 4 (review) | Check: accuracy vs stated claims, no unsourced assertions, BLUF is present and correct, all P0 coverage areas addressed, no inline provenance tags |
modules/wiki-placement.md | Starting Phase 5 (publish) | Place under the most relevant existing parent page. Confirm location with user before publishing. |
Two Entry Modes
| Mode | Trigger Signal | Entry Point |
| Proactive | "interview me about...", "capture my expertise..." | Phase 1 -> Phase 2 (full interview) |
| Reactive | "turn this into wiki docs...", "formalize this discussion..." | Phase 1 -> Phase 1.5 (harvest conversation) -> Phase 2 (gap-fill only) |
Phase 1: Scope (2-3 exchanges)
- Ask: What domain/topic? Who will read this? What type? (reference/runbook/architecture/onboarding)
- Confirm scope boundaries: in and out.
- New-vs-update check: Search your team's wiki for existing pages on this topic. If match: ask -- update, companion, or new?
- Determine entry mode: if the conversation already contains substantive domain content AND the user's trigger was reactive -> set
Source: conversation-> proceed to Phase 1.5. Otherwise -> setSource: interview-> proceed to Phase 2. - Load
modules/state-format.md-> create state file. Persist existing page ID and URL if update mode.
HARD GATE: Do not proceed without: topic, audience, intent, scope, entry mode, and new-vs-update decision.
Phase 1.5: Conversation Harvest (reactive mode only)
- Load
modules/coverage-matrix.md-> initialize coverage matrix in state file. - Scan the current conversation for domain-relevant content. Extract claims, decisions, examples, failure modes, terminology.
- For each claim: tag provenance as
[sme-stated](from user's words) or[inferred](agent synthesis). - Map extracted claims to coverage matrix. Mark areas as
covered,partial, oropen. - Append to interview log as harvested entries (
H<N>prefix, notQ<N>). - Present summary to user: "Here's what I extracted... These areas still have gaps: [list]."
- If all P0 areas covered -> Phase 2.5 (synthesis). If gaps remain -> Phase 2 (gap-fill).
HARD GATE: Do not skip this phase in reactive mode.
Phase 2: Interview (coverage-driven)
- Load
modules/coverage-matrix.mdif not already loaded (proactive mode). - Reactive mode shortcut: if Phase 1.5 covered all P0 areas and >=3 P1 areas, ask: "Proceed to synthesis or do you want follow-up questions?"
- Run question loop: one question at a time, driven by highest-priority uncovered area.
- The interviewee should experience a natural conversation, NOT a checklist.
- After each answer: update coverage, tag provenance, check for contradictions.
- Append Q&A to interview log (append-only,
Q<N>prefix).
HARD GATE: Do not proceed to Phase 2.5 without sufficiency gate passing OR explicit user stop.
Phase 2.5: Synthesize (agent-internal)
Extract claims from log. Cluster by coverage area. Surface contradictions (return to Phase 2 if unresolved). Preserve provenance. Map claims to article outline.
Phase 3: Draft
- Load
modules/bluf-template.md. - Generate article per template. Provenance goes in Source Notes appendix, NOT inline.
- Present draft to interviewee for initial reaction.
HARD GATE: Interviewee must see the draft before proceeding.
Phase 4: Review (min 2 rounds)
- Load
modules/review-rubric.md. - Dispatch sub-agent reviewer or self-review with explicit role switch.
- Present findings. Ask about gaps or disagreements.
- Fix -> re-review -> repeat. Min 2 rounds, max 3.
HARD GATE: 0 critical + 0 major findings before Phase 5.
Phase 5: Publish
- Load
modules/wiki-placement.md. - Run pre-publish checks: secret scan, duplicate detection.
- Run placement algorithm. Ask interviewee to confirm location.
- Ask for explicit publish approval.
- Publish. Verify. Update state file.
HARD GATE: Explicit interviewee approval required before publishing.
Resume
On "resume knowledge-capture": check state file in ~/.codex/knowledge-capture/. List topic, phase, source mode. Ask: resume or abandon? On abandon -> archive.
Full detail: references/scope-and-resume.md
Failure Modes
| Failure | Fix |
| Generic questions (not topic-specific) | Use coverage matrix. Ask about specifics, not "tell me more." |
| Checklist-feeling interview | Matrix is internal. Conversation should feel natural. |
| State file corruption on resume | Append-only log. Phase is single-field update. |
| Published article has inline provenance tags | Tags go in Source Notes appendix only. |
| Duplicate wiki page created | Run duplicate detection before publish. |
| Reactive mode skips harvest phase | Phase 1.5 is mandatory in reactive mode. |
| Wrong mode selected | Agent determines mode in Phase 1 based on trigger + conversation context. |
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.

