agentleFS
TheCollieryTheColliery/CoalLedger/skills/doc-consistency/SKILL.md

doc-consistency

Skill2 starsChanged 6 days agoApache-2.0View on GitHub

Docs-health contradiction scan—docs that disagree with EACH OTHER: the same fact stated two ways in two places, terminology drift (one thing under several names, or one name for several things), copy-paste blocks that were updated in one home and not the other, and cross-LANGUAGE drift (a doc and its counterpart in another language no longer say the same thing). Catches CONTRADICTORY doc sets. Triggers on: "/doc-consistency", "doc-consistency", "docs contradict", "terminology drift", "are the translations in sync". A mechanical layer (~free) checks declared translation pairs (`translationPairs`); the three drift classes are semantic (paid, consent-gated—comparing meaning across files is judgment, not pattern-matching). Severity judged by context, never mechanical; CONFIRMED contradictions and SUSPECTED tensions reported separately.

4 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions4 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 4

    1. AndrewBrown-USDA/aigb-skills-usda★ 7same name
    2. MaximSTH/meta-layer-starter★ 2same name
    3. wukun2005-gif/ai-skills★ 2same name
    4. dongsheng123132/doc-consistency★ 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 repository6 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    After any report in an interactive session you MUST present this menu via your question tool (skip only when findings are zero or no user …

    line 54
  2. Distinctive:

    NEVER auto-fix: every consistency fix is a content decision (which side wins)—there are no "safe" automatic fixes in this canary.

    line 60
  3. Very specific:

    If this canary misbehaves, OFFER to file it at <https://github.com/TheColliery/CoalLedger/issues> with a user-reviewed summary—never …

    line 66

≈ “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

Doc-Consistency

Answer in the USER'S language; keep technical terms, commands, paths, and check ids verbatim.

Find places where the doc set disagrees with itself. Report CONFIRMED contradictions; park tensions in SUSPECTED.

Parameters

  • SCOPE: a named doc set (default when given) | all docs touching a named topic | whole repo docs—.md/.mdx/.markdown/.rst/.txt/.adoc/.asciidoc/.org (confirm first if > 20 files).
  • TIER: two layers. Mechanical (~free): declared translation pairs only (Method 2b). Semantic (Full, paid, always consent-gated): the three drift classes below. Two sentences can contradict with zero textual overlap, so they stay judgment, never pattern-matching.

What counts (three drift classes)

classlooks like
fact driftthe same fact stated differently in two docs (a count, a default, a step order, a supported-platform claim)
terminology driftone concept under several names, or one name reused for different concepts—including a doc set's own defined terms used off-definition
cross-language drifta doc and its other-language counterpart diverging in meaning (a fact updated in one language only, a section present in one and absent in the other)

Method

  1. Inventory: extract the facts and defined terms each in-scope doc asserts (skip opinions; technical terms stay verbatim—a term is only "drifted" when the CONCEPT diverges, not when prose around it varies).
  2. Cross-compare the inventory; pair up disagreements. For cross-language pairs, compare MEANING (a free translation is fine; a contradicting one is drift).
    • 2b. Declared translation pairs (mechanical, ~free, SUSPECTED-only). Pairs come only from translationPairs (.coalledger.json, global + project merge), which the engine reads itself: never copy a pair into a command. Empty (factory default []): the engine reports nothing; never guess pairs from file names. The engine ships INSIDE this skill folder at ./lib/. Substitute your context's skill base dir for <skill base dir>. Run cd "<skill base dir>" && node ./lib/translation-pairs.mjs --list-path first, then write the run list with your file tool to the exact path it prints ({"root": "<absolute project folder>", "docs": []}), then run exactly: cd "<skill base dir>" && node ./lib/translation-pairs.mjs --json --list. The cd is REQUIRED (./lib/ resolves against the skill folder, never your project cwd). The run list's root is the absolute project folder; pair sides are relative to it, forward slashes, no ... No path goes on the command line; the engine reads the root from the run list as data.
      • Checks (all SUSPECTED): pair-invalid (no =, an empty side, a backslash, an absolute path or a .. segment; nothing is read) · pair-source-missing and translation-missing (a side is not a plain file under the root: missing, or refused by the link rule) · translation-older (the counterpart's last commit is before its source's) · translation-undated (no git record: no git, not a repository, or untracked or never committed; it means undated, never older).
      • Dates are each file's last git committer date (git log -1 --format=%cI), never mtime.
      • --json prints { "findings": [...] }. Exit 0 whenever the run completes, findings included; exit 2 on a usage error.
      • Which side is right stays out of this step (Escalation boundary).
  3. Which side is right is NOT this canary's call—report the pair and, where a source of truth is obvious, note it; resolving truth is doc-grounding's job (offer to run it on the pair).
  4. Severity by CONTEXT (never a fixed map), then honor severityFloor: contradictory instructions readers may follow = HIGH-CRITICAL; contradictory descriptions = MEDIUM; naming inconsistency = LOW. scanEverything: true bypasses the floor this run—report everything down to low—and say so: state that severityFloor was bypassed, never that every scope cut was bypassed (this canary has none to bypass).

Escalation boundary

Adjudicating WHICH of two contradicting claims is true = doc-grounding (fetch/recompute) or, for the error-not-allowed slice, CoalBoard (/coalboard) if that skill is installed. This canary only surfaces the disagreement.

Grants & denials (CLASSIFY-BLOCK)

classstep it powersgranton denial
readinventory facts/terms across the doc set, incl. cross-language pairsRead·Grep·Glob (·Bash·Write for the run list—the translation-pair engine runs via node ./lib/translation-pairs.mjs, Method 2b; Read/Grep/Glob cannot execute it)refuse that file, report it unscanned—never a false clean bill
writeAlign to a chosen side (after the user picks)Write·Edit (·Bash—checkpoint via git stash/commit)report + courier the intended change to the dispatcher; never claim applied

A denial reaches the WORKER as a visible message and propagates NO further—not to the dispatcher, not as a catchable condition. Every row above states a branch or an explicit death; a step that dies says so in the output. Never report a denied step as done, skipped, or clean.

Output

| # | doc A (path:line) | doc B (path:line) | class | disagreement | severity |

CONFIRMED table only; SUSPECTED (a tension that may be intended, e.g. an archive vs a live doc) as a separate list.

Reporting: call ReportFindings when callable—the schema carries ONE file/line pair per finding, so file/line MUST be doc A's defect site; doc B's location is cited as path:line inside summary/failure_scenario, never a paraphrase—one contradiction is one finding, never split across two. An unresolvable line reports your best guess, named imprecise in the wrap-up, never dropped. Severity prefixed in summary (e.g. [HIGH] …) per the severity-by-context rule above, ranked most-severe first, SUSPECTED tensions as verdict: PLAUSIBLE; chat then carries only the wrap-up line (counts · SUSPECTED list · overflow past 32) + the fix menu, never a restatement. Not callable → the table above, unchanged. No safe-fix class exists here—every fix is a which-side-wins content decision, so an Apply-fixes click degrades to the Let-me-pick path below, never an auto-apply. After any fix round, re-report the same findings with outcome: fixed/skipped/no_change_needed.

Fix mode (choice-gated)

After any report in an interactive session you MUST present this menu via your question tool (skip only when findings are zero or no user is present). NEVER auto-fix a live doc.

  • Align to a chosen side: the user picks which doc is right per finding; the other is edited to match, with a checkpoint first (git stash/commit in a git repo; else copy the file aside—never assume git exists).
  • Let me pick: list findings; the user selects.
  • Report only: exit unchanged.

NEVER auto-fix: every consistency fix is a content decision (which side wins)—there are no "safe" automatic fixes in this canary.

Multilingual

Cross-language comparison IS this canary's core case, worked in the languages the docs are written in. A poorly-handled language degrades to low-confidence SUSPECTED flags, never false alarms.

Problem report

If this canary misbehaves, OFFER to file it at https://github.com/TheColliery/CoalLedger/issues with a user-reviewed summary—never auto-submit.

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.