agentleFS
Sign inStart free
cachewraithcachewraith/cachewraith-skills/skills/ask/SKILL.md

ask

Skill0 starsChanged 8 days agoMITView on GitHub

Answer a question directly, after checking the real source of truth for it — the code, the config, the command output — instead of answering from memory. Deliberately cheap: a narrow lookup, then a short answer with the file:line that backs it. Invoke when the user asks a question about this project, this machine, or how something works and wants an answer rather than work. Skip when they want something built, changed, or committed, and skip when the question is open-ended enough to need a back-and-forth — that is discuss.

We split this skill into the rules it gives an agent

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

  • 0

    Common

    Most of the 58 edited versions we read keep these.

  • 3

    Distinctive

    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:

    Never answer a repo question from memory or from the file tree's shape.

    line 24
  2. Very specific:

    Do not: read whole files when a range answers it, run a broad find over the tree, open five files to confirm what the first one already …

    line 42
  3. Distinctive:

    Stop searching the moment you can answer.

    line 47
  4. Distinctive:

    Do not add: a restatement of the question, a description of how you searched, a summary of the answer you just gave, unrequested …

    line 64

“Common” is measured: kept by at least 6 in 10 of the 58 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

Ask

Answer the question. Nothing else.

Two rules carry this skill, and they pull against each other on purpose: check before you answer, and spend as little as possible doing it.

1. What kind of question is this

Decide before touching a tool:

QuestionWhat to do
The answer lives in this repo (what does X do, where is Y, is Z wired up)Look it up. One targeted search, read only the lines that matter.
The answer lives on the machine (versions, branch, what is installed, why a command failed)Run the one command that shows it.
General knowledge you actually hold, with no project-specific partAnswer straight, no tools.
The answer changed recently or depends on the outside world (a library's current API, a release, a CVE)Say you'd have to look it up, and do one search — not five. No web search tool (Codex without --search, say)? Say what you'd look up instead of guessing.

Never answer a repo question from memory or from the file tree's shape. If you have not seen the line, you do not know it.

2. Look, narrowly

Budget: three tool calls. Four or five when the question genuinely has parts. If you are past that and still guessing, stop and say what you would need to check next — a wrong answer delivered confidently costs the user more than a short "I'd need to look at X".

Cheap moves, in order of preference:

rg -n 'PATTERN' --glob '!node_modules' -m 20        # locate, don't dump
rg -n 'PATTERN' -l                                  # just the filenames first
sed -n '40,80p' path/to/file                        # read the range, not the file
git log --oneline -5 -- path/to/file                # when the question is "why"

Do not: read whole files when a range answers it, run a broad find over the tree, open five files to confirm what the first one already said, or spawn a subagent. A subagent starts cold and re-derives everything you already have — for one question it is the most expensive way to get the cheapest answer.

Stop searching the moment you can answer. Confirmation beyond that is spending.

3. Answer

Lead with the answer. First sentence, no preamble.

  • Length matches the question. A yes/no question gets a yes or no plus the reason. A "where is X" gets file:line. A "how does X work" gets a short paragraph or a few bullets — not a tour of the module.
  • Cite what you checked — src/auth/session.ts:42 — so the user can verify in one click. One or two citations, the load-bearing ones, not a bibliography.
  • Separate seen from inferred. "The handler validates the token at auth.ts:31" is something you read. "So an expired token would 401 here" is inference — mark it as such when it matters.
  • Say when you don't know. "Not in this repo — it's set by the deploy env" is a real answer. Guessing is not.

Do not add: a restatement of the question, a description of how you searched, a summary of the answer you just gave, unrequested suggestions, or an offer to implement it. If the answer implies obvious next work, one short line at the end is the whole allowance.

4. Do not act

This skill reads. It does not edit, create, commit, install, or run anything with a side effect. If the answer is "that's broken", say it's broken and stop — the user asked what, not fix it. They will ask for the fix if they want it.

Download now (ZIP)

  • Other versions70 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 70

    1. Yeachan-Heo/oh-my-claude-sisyphus★ 40ksame name
    2. Yeachan-Heo/oh-my-claudecode★ 40ksame name
    3. Yeachan-Heo/oh-my-codex★ 34ksame name
    4. agenticnotetaking/arscontexta★ 3.5ksame name
    5. acunniffe/git-ai★ 2.9ksame name
    6. git-ai-project/git-ai★ 2.9ksame name
    7. DAIR-AutoResearch/AutoSci★ 1.7ksame name
    8. PKU-DAIR/AutoSci★ 1.7ksame name
    9. skyllwt/AutoSci★ 1.7ksame name
    10. skyllwt/OmegaWiki★ 1.7ksame name

    60 more not listed

    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.

  • Used alongsideProjects with it also use daily-arxiv, exp-eval and 10 more.

    Other files people use alongside this one.

    Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
    • LambenthanSkill

      daily-arxiv

      每日从 arXiv 拉取新论文，过滤相关性，auto-ingest 高优先级论文，检测 SOTA 更新

      Also in 5 projects that use this

    • LambenthanSkill

      exp-eval

      实验判决门：Review LLM 独立评判实验结果 → 4 种判决路径 → 自动更新 claims confidence、ideas status、graph edges

      Also in 5 projects that use this

    • DAIR-AutoResearchSkill

      exp-pilot-eval

      Pilot result evaluation — read pilot results, apply success criteria, update idea page (pilot_result, failure_reason if failed), generate PILOT_VERDICT_REPORT. Called by /ideate Phase 5 after /exp-pilot-run.

      Also in 4 projects that use this

    • DAIR-AutoResearchSkill

      exp-pilot-run

      Pilot experiment execution — read Pilot Spec YAML, write pilot code, run experiment(Confirm with the user before operation and require the applicant to conduct manual inspection), return results. Called by /ideate Phase 5. Does NOT modify wiki pages or judge pass/fail.

      Also in 4 projects that use this

    • phanghonghaoSkill

      poster

      HTML/CSS A4 海报生成。从沙龙/活动 Markdown 文件生成科技风 A4 海报 PDF/PNG。Use when user mentions: poster, 海报, 海报生成, make poster, generate poster, 活动海报, 沙龙海报.

      Also in 4 projects that use this

    • +6 more from Lambenthan
    • +1 more from DAIR-AutoResearch
  • In this repository11 other files it gives its agents.

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

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.

Check your setup, free