agentleFS
flonatflonat/flonat-research/skills/knowledge-lint/SKILL.md

knowledge-lint

Skill150 starsChanged 9 days agoMITView on GitHub

Lint an existing project knowledge base for contradictions, uncited claims, missing connections, stale articles, and orphaned concepts. Use when checking the integrity of already compiled knowledge before relying on or expanding it. Not for compiling raw sources; use $compile-knowledge.

The file

We split this skill into the rules it gives an agent

10 rules. About 0 are common, 3 are distinctive, and 7 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 7

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Idempotency: if today's file exists, append a same-day descriptor ({date}-revision.md, {date}-r2.md, {date}-pre-submission.md) — never …

    line 20
  2. Very specific:

    Index update: if reviews/INDEX.md exists, write a one-line entry under "Latest per source" pointing at the new file.

    line 21
  3. Very specific:

    After running compile-knowledge to verify integrity

    line 27
  4. Distinctive:

    Before writing a paper section to check the knowledge base is solid

    line 28
  5. Distinctive:

    When switching between projects to verify knowledge is current

    line 30
  6. Very specific:

    If no knowledge/ directory exists — run compile-knowledge first

    line 34
  7. Very specific:

    Verify knowledge/ exists and has at least 2 articles (linting a single article is pointless)

    line 42
  8. Very specific:

    If missing: "No knowledge directory found.

    line 43
  9. Distinctive:

    Check for concepts that should link to each other but don't:

    line 79
  10. Very specific:

    Write correspondence/internal-reviews/KNOWLEDGE-LINT-REPORT.md (create directory with mkdir -p if needed):

    line 125

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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.

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

Knowledge Lint

Substantive consistency checks on the compiled knowledge wiki. Produces correspondence/internal-reviews/KNOWLEDGE-LINT-REPORT.md.

Output Path

Per rules/review-artefact-routing.md (auto-loads in research projects (path-scoped to paper-*/ and paper/)):

  • Source slug: knowledge-lint
  • Write reports to: reviews/_project/knowledge-lint/YYYY-MM-DD-HHMM.md inside the project. Path is relative to the research project root, not the Task-Management repo.
  • Never at project root (./CRITIC-REPORT.md-style filenames are forbidden — pre-rule layout).
  • Idempotency: if today's file exists, append a same-day descriptor ({date}-revision.md, {date}-r2.md, {date}-pre-submission.md) — never overwrite.
  • Index update: if reviews/INDEX.md exists, write a one-line entry under "Latest per source" pointing at the new file. Otherwise review-recap will rebuild the index next time it runs.
  • Infrastructure repos (Task-Management, atlas-workspace, etc.): this section does not apply — the path-scoped rule won't load there.

When to Use

  • After running compile-knowledge to verify integrity
  • Before writing a paper section to check the knowledge base is solid
  • Periodically (monthly) to catch drift
  • When switching between projects to verify knowledge is current

When NOT to Use

  • If no knowledge/ directory exists — run compile-knowledge first
  • For structural project audits — use an installed research-project audit workflow
  • For bibliography validation — use the configured bibliography validator

Pre-Check

  1. Verify knowledge/ exists and has at least 2 articles (linting a single article is pointless)
  2. If missing: "No knowledge directory found. Run compile-knowledge first."

Checks

1. Contradictions

Scan all articles for claims about the same concept that disagree:

CheckHow
Same concept, different claimsCompare Key Findings sections across articles that share wikilinks
Method disagreementsCheck if two articles recommend different approaches for the same problem
Factual conflictsCross-reference dates, numbers, definitions across articles

Report format:

CONTRADICTION: concept-a.md says "X increases Y" but concept-b.md says "X has no effect on Y"
  Source A: Author (2023) — concept-a.md line 15
  Source B: Author (2024) — concept-b.md line 22
  Resolution needed: check which paper's context applies

2. Uncited Claims

Scan Key Findings sections for assertions without source attribution:

CheckHow
Missing citationsFindings that don't reference a paper, session log, or decision
Vague citations"Some studies show..." without specific references
Self-referentialCiting another knowledge article instead of a primary source

3. Missing Connections

Check for concepts that should link to each other but don't:

CheckHow
Shared keywordsTwo articles mention the same method/dataset/author but don't cross-reference
Atlas connectionsAtlas topic has connected_topics that don't have corresponding knowledge articles
Orphaned articlesArticles with zero incoming wikilinks from other articles

4. Staleness

CheckHow
Stale articlesLast updated > 90 days ago AND project has had session activity since
Source driftNew papers in docs/literature-review/ not compiled into any article
Memory driftNew [LEARN] entries in MEMORY.md not reflected in knowledge articles
Session driftRecent session logs mention findings not in any knowledge article

5. Coverage Gaps

CheckHow
Key paper not compiledPapers cited in paper-*/paper/*.tex but not in any knowledge article
Method not documentedEstimators/methods used in code but not explained in knowledge
Concepts mentioned but not definedTerms used across multiple articles without a dedicated article

6. Promotion readiness (advisory)

Project knowledge articles are the upstream source that wiki-grow promotes into vault concepts (~/vault/concepts/). Articles that escape their origin project (mentioned in ≥3 corpus docs) get auto-promoted as status: draft and later curated to the concept anatomy (packages/research-atlas-schema/schema.md → "Concept File Schema"): a lead ## In one line definition, encyclopedia-voice body, and a ## In my portfolio section for project-specific prose.

This is advisory, not a hard check — project knowledge lives in the project's own voice and needn't conform. But flagging an article that leads with a crisp, project-agnostic definition as "promotion-ready" (vs. one that opens mid-analysis) helps: the cleaner the lead, the cheaper the eventual draft→curated pass. Note such articles in Recommendations; don't rewrite them here.


Report

Write correspondence/internal-reviews/KNOWLEDGE-LINT-REPORT.md (create directory with mkdir -p if needed):

# Knowledge Lint Report

**Project:** [name]
**Date:** YYYY-MM-DD
**Articles scanned:** N
**Total words:** ~X

## Summary

| Check | Issues | Severity |
|-------|--------|----------|
| Contradictions | N | High |
| Uncited claims | N | Medium |
| Missing connections | N | Low |
| Staleness | N | Medium |
| Coverage gaps | N | Medium |

## Contradictions (High)

[List each with source references and suggested resolution]

## Uncited Claims (Medium)

[List each with article and line]

## Missing Connections (Low)

[List pairs that should link]

## Staleness (Medium)

[List stale articles with last-updated date and evidence of new activity]

## Coverage Gaps (Medium)

[List uncompiled sources and undocumented concepts]

## Recommendations

1. [Most impactful fix]
2. [Second]
3. [Third]

Cross-References

SkillRelationship
compile-knowledgeRun this first to build/update the wiki
store-insightUse to fix individual gaps found by lint
research-atlas-coherenceComplementary — checks topic-level connections, not article-level
wiki-curateDownstream — audits the vault concepts these articles promote into (anatomy, lifecycle, overlap)

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.