agentleFS
haruharu/redmine_ai_helper/.claude/skills/speckit-wiki-query/SKILL.md

speckit-wiki-query

Skill107 starsChanged 9 months agoMITView on GitHub

Answer a question from the wiki with page and source citations; flag

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are 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.

  • 3

    Very specific

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

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

    Read INDEX.md and pick the pages most relevant to the question — at most query.pages_slice, staying within the query.context_tokens render …

    line 43
  2. Very specific:

    Prefer pages whose type matches the question's shape (a "why did we…" question → decision pages.

    line 45
  3. Distinctive:

    Read only the selected pages.

    line 51
  4. Distinctive:

    every load-bearing statement names its wiki page and carries the page's source IDs, e.g.

    line 53
  5. Very specific:

    If pages disagree, report the disagreement.

    line 78

≈ “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.

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

Wiki Query Skill

Query the Wiki

Answer a question from the wiki pages, with citations — or say plainly that the wiki cannot answer it yet and what to ingest to fix that. This is the query operation of Karpathy's LLM Wiki pattern, and it is also how you test the wiki: an answer the pages cannot support is a coverage gap, not an invitation to improvise.

User Input

$ARGUMENTS

$ARGUMENTS is the question. If empty, give a one-screen overview of what the wiki currently knows (scope, page types and counts, notable pages) and stop.

Steps

1. Resolve the wiki

Load configuration exactly as /speckit.wiki.init does. If WIKI_DIR/SCHEMA.md does not exist, report that no wiki exists and recommend /speckit.wiki.init — do not answer the question from general knowledge.

2. Select pages

Read INDEX.md and pick the pages most relevant to the question — at most query.pages_slice, staying within the query.context_tokens render budget. Prefer pages whose type matches the question's shape (a "why did we…" question → decision pages; a "how does X work" question → component pages).

3. Answer with citations

Read only the selected pages. Compose the answer such that:

  • every load-bearing statement names its wiki page and carries the page's source IDs, e.g. …retries are idempotent by key (idempotency-keys, S003);
  • conflicting page content (> ⚠ conflict: markers) is surfaced as conflicting, with both sides and their sources — never pick a winner silently;
  • pages the answer relied on are listed at the end.

4. State coverage honestly

Close with one of:

  • Covered — the pages fully support the answer.
  • Partial — say exactly which part is unsupported, and recommend the concrete ingest that would close the gap (/speckit.wiki.ingest <likely source>).
  • Uncovered — the wiki has nothing; say so, and recommend where the answer likely lives (a feature's research.md, a file, a URL) as an ingest target.

Guardrails

  • Read-only. This command writes nothing — no pages, no index, no registry.
  • Answer only from wiki pages. General knowledge may be used to phrase the answer, never to supply facts the pages do not contain.
  • If pages disagree, report the disagreement; resolving it is /speckit.wiki.lint's and the user's job.
  • Respect the slice and token caps — a focused answer from 5 pages beats a vague one from 50.

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.