agentleFS
arthur-plautzarthur-plautz/research-toolkit/.claude/skills/update-project-knowledge/SKILL.md

update-project-knowledge

Skill0 starsChanged 4 months agoMITView on GitHub

Interview the user to fill in or correct the project knowledge docs (docs/project/overview.md, architecture.md, analysis.md, research.md). Use when these docs are still placeholder/templated, out of date, or the user wants to walk through setting them up or updating them.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

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

    Before starting, read all four docs to see what's already filled in vs.

    line 10
  2. Very specific:

    Follow /grilling's method exactly: one question at a time, give a recommended answer with each question, wait for feedback before …

    line 14

≈ “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

Same mechanism as /grill-with-docs, retargeted: a /grilling session whose output is docs/project/overview.md, architecture.md, analysis.md, and research.md instead of /domain-modeling's CONTEXT.md/ADRs.

Scope

Before starting, read all four docs to see what's already filled in vs. placeholder ({...} text). Ask only about what's missing, placeholder, or flagged as stale — don't re-litigate sections that already hold real content unless the user raises them.

Interview

Follow /grilling's method exactly: one question at a time, give a recommended answer with each question, wait for feedback before continuing. If a question is answerable by reading the code, explore instead of asking.

Walk the docs in this order — each depends on answers from the ones before it:

  1. overview.md — what the project is, its components/experiment types, mode/variant switches
  2. architecture.md — class/module structure, ownership chain, data flow
  3. analysis.md — experiment types, data output format, key metrics, conventions
  4. research.md — the current research question, rationale, references

Write inline

The moment a question resolves, write the answer straight into the relevant doc section — don't batch edits to the end of the session. Replace placeholder text ({...}) with the resolved content; never leave both the placeholder and the answer sitting in the same section.

research.md claims that aren't common knowledge need grounding: invoke /references before writing them down, per .claude/rules/references.md.

End of session

Once all four docs are filled or the user ends the interview, summarize what changed in one or two lines — do not restate the docs' full content back.

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.