agentleFS
bordenetbordenet/superpowers-plus/skills/productivity/knowledge-capture/skill.md

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

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
  1. Very specific:

    Ask: What domain/topic? Who will read this?

    line 66
  2. Distinctive:

    New-vs-update check: Search your team's wiki for existing pages on this topic.

    line 68
  3. Very specific:

    Determine entry mode: if the conversation already contains substantive domain content AND the user's trigger was reactive -> set Source: …

    line 69
  4. Distinctive:

    HARD GATE: Do not proceed without: topic, audience, intent, scope, entry mode, and new-vs-update decision.

    line 72
  5. Distinctive:

    HARD GATE: Do not skip this phase in reactive mode.

    line 86
  6. Distinctive:

    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?"

    line 93
  7. Distinctive:

    Run question loop: one question at a time, driven by highest-priority uncovered area.

    line 94
  8. Distinctive:

    After each answer: update coverage, tag provenance, check for contradictions.

    line 96
  9. Distinctive:

    HARD GATE: Do not proceed to Phase 2.5 without sufficiency gate passing OR explicit user stop.

    line 99
  10. Distinctive:

    Run pre-publish checks: secret scan, duplicate detection.

    line 133
  11. Distinctive:

    Run placement algorithm. Ask interviewee to confirm location.

    line 134
  12. Distinctive:

    Ask for explicit publish approval.

    line 135

“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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
skill.md

knowledge-capture

Announce at start: "I'm using the knowledge-capture skill to capture expertise and produce a wiki article."

Routing

SignalWrong SkillRight Skill
"interview me about..."wiki editingknowledge-capture
"turn this conversation into a wiki article"brainstormingknowledge-capture
"edit this wiki page"knowledge-capturewiki editing
"write a design doc"knowledge-capturebrainstorming

Collision note: expert-interviewer shares "interview me about". knowledge-capture takes precedence — it is the heavier-duty, wiki-integrated version.


Module Loading (MANDATORY)

ModuleLoad BEFOREIf unavailable, use:
modules/state-format.mdAny state file read/writeTrack state in a Markdown file with sections: Topic, Phase, Source (interview/conversation), Wiki Page ID, Interview Log (append-only Q&A)
modules/coverage-matrix.mdStarting Phase 1.5 or Phase 2Use areas: Context/Background, Core Concepts, How it Works, Failure Modes, Common Mistakes, Expert Tips, References. Mark each: open/partial/covered.
modules/bluf-template.mdStarting Phase 3 (drafting)Article format: ## Bottom Line (1-2 sentences), ## Background, ## How It Works, ## Failure Modes, ## References + Source Notes
modules/review-rubric.mdStarting 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.mdStarting Phase 5 (publish)Place under the most relevant existing parent page. Confirm location with user before publishing.

Two Entry Modes

ModeTrigger SignalEntry 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)

  1. Ask: What domain/topic? Who will read this? What type? (reference/runbook/architecture/onboarding)
  2. Confirm scope boundaries: in and out.
  3. New-vs-update check: Search your team's wiki for existing pages on this topic. If match: ask -- update, companion, or new?
  4. 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 -> set Source: interview -> proceed to Phase 2.
  5. 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)

  1. Load modules/coverage-matrix.md -> initialize coverage matrix in state file.
  2. Scan the current conversation for domain-relevant content. Extract claims, decisions, examples, failure modes, terminology.
  3. For each claim: tag provenance as [sme-stated] (from user's words) or [inferred] (agent synthesis).
  4. Map extracted claims to coverage matrix. Mark areas as covered, partial, or open.
  5. Append to interview log as harvested entries (H<N> prefix, not Q<N>).
  6. Present summary to user: "Here's what I extracted... These areas still have gaps: [list]."
  7. 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)

  1. Load modules/coverage-matrix.md if not already loaded (proactive mode).
  2. 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?"
  3. Run question loop: one question at a time, driven by highest-priority uncovered area.
  4. The interviewee should experience a natural conversation, NOT a checklist.
  5. After each answer: update coverage, tag provenance, check for contradictions.
  6. 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

  1. Load modules/bluf-template.md.
  2. Generate article per template. Provenance goes in Source Notes appendix, NOT inline.
  3. Present draft to interviewee for initial reaction.

HARD GATE: Interviewee must see the draft before proceeding.


Phase 4: Review (min 2 rounds)

  1. Load modules/review-rubric.md.
  2. Dispatch sub-agent reviewer or self-review with explicit role switch.
  3. Present findings. Ask about gaps or disagreements.
  4. Fix -> re-review -> repeat. Min 2 rounds, max 3.

HARD GATE: 0 critical + 0 major findings before Phase 5.


Phase 5: Publish

  1. Load modules/wiki-placement.md.
  2. Run pre-publish checks: secret scan, duplicate detection.
  3. Run placement algorithm. Ask interviewee to confirm location.
  4. Ask for explicit publish approval.
  5. 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

FailureFix
Generic questions (not topic-specific)Use coverage matrix. Ask about specifics, not "tell me more."
Checklist-feeling interviewMatrix is internal. Conversation should feel natural.
State file corruption on resumeAppend-only log. Phase is single-field update.
Published article has inline provenance tagsTags go in Source Notes appendix only.
Duplicate wiki page createdRun duplicate detection before publish.
Reactive mode skips harvest phasePhase 1.5 is mandatory in reactive mode.
Wrong mode selectedAgent determines mode in Phase 1 based on trigger + conversation context.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.