agentleFS
inclusionAIinclusionAI/Choruz/.agents/skills/choruz-doc/SKILL.md

choruz-doc

Skill1.1k starsChanged 18 days agoApache-2.0View on GitHub

Create, restructure, review, audit, or migrate Choruz Markdown documentation (docs/, README, AGENTS.md, the in-app docs pages) using one owner per fact, tier placement, executed-operation fact-checking, current-state prose, and the repository validators. Use for new or revised docs, docs-tree organisation, and documentation-quality audits.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    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. Distinctive:

    Read AGENTS.md, the target page, its source or tests, and the page that links to it.

    line 31
  2. Very specific:

    Place it at its nearest owner from the table.

    line 33
  3. Distinctive:

    Write the common path briefly and link exhaustive truth one hop away.

    line 35
  4. Very specific:

    Verify every claim against code, tests, migrations, scripts, or package.json, and run the operations the page instructs (below).

    line 36
  5. Distinctive:

    Run the validators, then re-read the whole diff once for correctness and once for brevity.

    line 37
  6. Distinctive:

    Delete what you could not reproduce: Never carry a command, field, default or behaviour from memory or from a neighbouring page.

    line 44
  7. Very specific:

    Check old pages against origin/main: before revising them; a stale statement on main is still wrong and is corrected against the code, …

    line 45
  8. Very specific:

    Use choruz-prose-standard for sentence-level judgement.

    line 77

≈ “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

Choruz documentation

Summary

Make every page findable, newcomer-readable, and exact enough for agents and maintainers, with one owner per fact. Apply AGENTS.md and the gates first, then this workflow for placement, progressive detail, fact-checking, and validation. Design rationale lives in Agent Notes; this skill owns everything else written for a human or agent reader.

The tiers: one home per fact

TierHomeHoldsNot
Standing ordersAGENTS.mdOne to three lines per rule, linking its homeRationale, procedures
Platform protocolCLAUDE.mdHow an agent talks to humans and agents on ChoruzEngineering rules
Architecturedocs/architecture.mdThe ordered map of how the system works todayPer-crate detail, history, status
Data modeldocs/data-model.mdTables, ownership, workspace_id, sequencesMigration narration
Operationsdocs/operations/Install, run, deploy, back up, SLO, runbookDesign rationale
Testingdocs/testing/PR types, what CI runs, e2e conventionsTest walkthroughs
Decisions.agents/notes/Problem, decision, alternatives, consequencesCurrent-state reference
Contributor entryCONTRIBUTING.md, README.mdSetup, commands, layout, links into the tiersDuplicated rules
User docsapps/web/app/docs/The in-app documentation siteEngineering process
Package contractsa package's README.md, doc commentsConfig, semantics, failures, limitationsCross-package narrative

Placement: bugs and incidents to a note or a runbook; rationale to notes; procedures to operations; type and table definitions to the data model; package contracts to the package README; standing orders to AGENTS.md with a link.

Workflow

  1. Read AGENTS.md, the target page, its source or tests, and the page that links to it.
  2. Classify the page by one job and reader: setup tutorial, operator procedure, architecture overview, reference, decision record, user guide.
  3. Place it at its nearest owner from the table. Keep package contracts beside the package; use docs/ for cross-package material.
  4. Define the reader's starting state, observable outcome, likely failure, recovery path and next depth before writing details.
  5. Write the common path briefly and link exhaustive truth one hop away. Each section opens with a short orienting paragraph.
  6. Verify every claim against code, tests, migrations, scripts, or package.json, and run the operations the page instructs (below).
  7. Run the validators, then re-read the whole diff once for correctness and once for brevity.

Fact-check procedure: test, do not assume

The only admissible evidence for an operation claim is having run it.

  1. Run every claimed command against the current checkout, exactly as the page will show it; write down only what you observed, including warnings and failure modes. pnpm scripts come from package.json; a documented flag must exist in the binary's --help.
  2. Delete what you could not reproduce. Never carry a command, field, default or behaviour from memory or from a neighbouring page. When a claim fails to reproduce, fix the claim.
  3. Check old pages against origin/main before revising them; a stale statement on main is still wrong and is corrected against the code, not against the old prose.
  4. Environment variables are named from their reading site (git grep CHORUZ_ over infra/, services/, apps/), with the default the code applies.

Voice rules

  • Summary says what the subject does for the reader (outcomes, when to choose it, main cost), never its role or internal identity.
  • Reference sections explain, never enumerate what a generated or source artifact already lists; link the source for exact detail.
  • Current state only. No "previously", "now", "no longer", PR numbers or migration narration outside notes' Consequences.
  • Controlled technical English: one actor and one main action per sentence where ambiguity could change behaviour; one term per concept; direct verbs; preserve modality and exceptions. Chinese is welcome in README.md and user-facing pages where the existing page is Chinese; keep a page in one language.
  • Link by relative path, never by bare filename or note number.

Quality criteria

  • Brief: the common path contains only facts needed for its outcome.
  • Intuitive: prerequisites precede dependent concepts; one next action is obvious; headings use terms readers search for.
  • Friendly: readers can recognise success, understand risk before acting, and recover from likely failure.
  • Accurate: each durable claim has one owner and a verification path.
  • Agent-readable: stable headings, anchors, terminology and ownership support targeted retrieval.
  • Newcomer-complete: an engineer with no repository context can reconstruct the relevant architecture through three to five linked pages.

Audit the corpus

Hunt the slop list with the cheapest probes first: git grep distinctive phrases to find duplicated rules; find hand-written inventories that a script or the tree already owns; find migration plans and future-tense spec language in implemented notes; find "previously/now/no longer". Keep every load-bearing rule, preferably as one to three lines plus a link to its rationale.

Validation

  • Links: every relative link resolves; check the targets or open the page in the rendered preview.
  • Agent Notes: python3 scripts/verify_agent_notes.py.
  • Commands: re-run every command the page instructs before merging a claim about it.
  • Web docs pages (apps/web/app/docs): pnpm web:check, and the docs e2e spec when the page structure changed.
  • A docs-only pull request runs no test job and merges in seconds; do not mix a config or code change into it.

Use choruz-prose-standard for sentence-level judgement.

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.