agentleFS
Sign inStart free
mad01mad01/thismoon/skills/ask/SKILL.md

ask

Skill2 starsChanged 10 days agoBSD-3-ClauseView on GitHub

Answer a question from the brain vault, the local markdown knowledge base, and cite the notes behind the answer. Use when the user says "/ask", "ask the brain", "what do we know about", "what do my notes say", "have I written about", or wants an answer from saved knowledge before anything is looked up elsewhere. Read-only. Backed by the brain MCP (brain_hybrid_search, brain_search, brain_peek, brain_read, brain_backlinks) with the `brain` CLI as fallback.

75 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects 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

  • skyllwtSkill

    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

  • skyllwtSkill

    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 skyllwt

The file

We split this skill into the rules it gives an agent

12 rules. 0 are common, 8 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 62 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    Prefer the MCP tools: brain_hybrid_search, brain_search, brain_peek, brain_read, brain_backlinks.

    line 19
  2. Very specific:

    Find the notes: Run brain_hybrid_search with the question in the user's words.

    line 27
  3. Very specific:

    Prefer notes/ and MOCs/ over inbox/, and a note with a higher evidence grade over a lower one.

    line 41
  4. Distinctive:

    Answer: Follow the reply rules below.

    line 48
  5. Distinctive:

    Every read tool returns the fields that say how far to trust a note.

    line 52
  6. Very specific:

    Search leaves out revoked notes (superseded, retracted, withdrawn).

    line 65
  7. Distinctive:

    Keep what the notes say apart from what you add.

    line 80
  8. Distinctive:

    When notes disagree, show both claims with their grades and dates.

    line 82
  9. Distinctive:

    When the vault has no answer, say that in the first line.

    line 84
  10. Distinctive:

    Do not write, edit or promote a note from this skill.

    line 89
  11. Distinctive:

    Do not fill a gap in the notes with a guess that reads like a citation.

    line 90
  12. Distinctive:

    Do not read a whole vault to answer one question.

    line 91

“Common” is measured: kept by at least 6 in 10 of the 62 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

Answers a question from the brain vault and shows which notes the answer comes from. The vault is the source. Nothing in the answer comes from the web or from model memory unless the reply says so.

ask only reads. It never writes, edits or promotes a note. To add knowledge, capture it with brain capture or brain_write. To research a topic the vault does not cover, say so and let the user choose the next step.

Tools

Prefer the MCP tools: brain_hybrid_search, brain_search, brain_peek, brain_read, brain_backlinks. The brain CLI is the fallback when the MCP server is not connected (brain hybrid-search, brain search, brain peek, brain read, brain backlinks). On a tool error or an empty result that looks wrong, call brain_doctor before a retry.

Steps

  1. Find the notes. Run brain_hybrid_search with the question in the user's words. It joins a literal search and a vector search. When its warnings[] say the vector half did not run, the hits are literal only: follow up with brain_search on one or two distinct terms from the question. When it returns an error, do the same and tell the user in one line that the vector search is down. A brain_search query is zoekt syntax: two words mean both must be in the same note, and a|b means either.
  2. Widen once on a miss. Try a synonym, the singular form, or a broader term. Search covers every vault and every tier, so do not narrow with tier or vault unless the user named one. After two misses stop and report that the vault holds nothing on the question.
  3. Pick what to read. brain_peek the top hits, five at most. Peek is cheap: it returns the frontmatter, the first lines and the backlink count. Prefer notes/ and MOCs/ over inbox/, and a note with a higher evidence grade over a lower one.
  4. Read the notes that answer. brain_read each one, with start_line and end_line when only a section matters. A map of content (a note under MOCs/) is an index: follow its wikilinks to the notes it lists. When a note points at a neighbour that looks closer to the question, check brain_backlinks and read that one too.
  5. Answer. Follow the reply rules below.

What the vault says about trust

Every read tool returns the fields that say how far to trust a note. Carry them into the answer.

FieldValueWhat to do
evidenceA, BState the claim plainly
evidenceCState the claim and say the evidence is weak
evidenceDThe claim is refuted. Say so. Never give it as the answer
evidencemixedGrades sit beside each claim in the body. Use the grade of the claim you quote
statusrawAn unchecked capture from inbox/. Say it is unverified
statusdisputedShow the banner from brain_read before the claim
staletrueThe note is past its review_by date. Say it may be out of date

Search leaves out revoked notes (superseded, retracted, withdrawn). Keep it that way: set include_revoked only when the user asks what the vault used to say.

The reply

  • Lead with the answer in one or two sentences.
  • Then the sources, one line each: the note title as a link, its evidence grade when it has one, and a flag for raw, disputed or stale.
  • Link a note to the brain web view when the machine runs it: the root URL plus ?note= and the URL-encoded note ref in the vault:path form. With the host brain.this, the note personal:notes/ops/k8s-upgrades.md is http://brain.this/?note=personal%3Anotes%2Fops%2Fk8s-upgrades.md. Without the web view, give the vault:path ref.
  • Quote a note for a number, a date or a decision. Paraphrase the rest.
  • Keep what the notes say apart from what you add. Mark an inference as yours.
  • When notes disagree, show both claims with their grades and dates. Do not pick a side without saying why.
  • When the vault has no answer, say that in the first line. Offer the general answer as a separate, labelled part only when the user wants it.

Do not

  • Do not write, edit or promote a note from this skill.
  • Do not fill a gap in the notes with a guess that reads like a citation.
  • Do not read a whole vault to answer one question. Five notes is the cap; ask the user to narrow the question past that.

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.

Check your setup, free