agentleFS
mikejhillmikejhill/agent-skills/skills/project-documentation/SKILL.md

project-documentation

Skill0 starsChanged 8 months agoMITView on GitHub

Organize and review project documentation, docs directories, README indexes, authority, and lifecycle metadata. Use for documentation conventions and profiles; technical-writing governs wording in any location.

6 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 5 edited versions we read keep these.

  • 3

    Distinctive

    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:

    Use root README.md as the project entrance.

    line 13
  2. Very specific:

    Every documentation directory, including docs/, has README.md containing its primary area content and a relative-link index of its direct …

    line 14
  3. Distinctive:

    Keep a topic in its directory README while it serves the area's main purpose.

    line 15
  4. Distinctive:

    Create reserved directories only when needed.

    line 16
  5. Very specific:

    Read references/metadata.md when authoring or reviewing content metadata.

    line 31
  6. Distinctive:

    Keep accepted decision reasoning historical.

    line 37

“Common” is measured: kept by at least 6 in 10 of the 5 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

Project documentation

Make knowledge discoverable, its authority explicit, and its maintenance affordable. This is the convention layer between project principles and scoped profiles.

Structure and discovery

  • Inspect the existing documentation and profiles first. Prefer Markdown and lowercase-kebab-case document names; retain conventional names and paths required by tooling.
  • Use root README.md as the project entrance. Put deeper knowledge in docs/, with topic directories only when content warrants them.
  • Every documentation directory, including docs/, has README.md containing its primary area content and a relative-link index of its direct child documents/directories. A README is useful content, not merely a link list.
  • Keep a topic in its directory README while it serves the area's main purpose. Split detail when it has a distinct audience, lifecycle, owner, procedure, or enough detail to obscure orientation. Link it from the README.
  • Create reserved directories only when needed. Their meanings are stable; profiles choose the exact tree and may add categories:
DirectoryReserved meaning
decisions/Decision rationale and lifecycle records
architecture/Current system structure and relationships
guides/Task-oriented usage and learning
reference/Authoritative lookup material and contracts
operations/Operating procedures, recovery, and operational orientation
plans/Exploratory proposals and intended work, not current requirements

An operations README gives orientation and canonical links. Link technical asset ownership, business ownership, and identifiers to their authoritative inventory; do not maintain another manual copy.

Trust and lifecycle

Read references/metadata.md when authoring or reviewing content metadata. Use references/metadata.schema.json to validate parsed YAML frontmatter. Apply it to content-bearing READMEs too.

  • Choose one canonical authority for each fact. Link to it rather than copying facts manually across pages.
  • A deterministic generated region may reproduce sourced facts if it labels its generator, source, and regeneration method. It is a derivative view, never a second authority. Keep hand-edited content outside the generated region.
  • Distinguish current requirements, observed behavior, historical decisions, and exploratory intent. Presentation quality, generation, or an accepted decision alone does not prove implementation or verification.
  • Record verification only after checking evidence, by the identified reviewer, for the stated scope. Never fabricate a reviewer or mark content verified because a model wrote it.
  • Keep accepted decision reasoning historical. Supersede it with a linked new decision when the choice changes. Preserve the old rationale; correct factual errors transparently.
  • Review content when its source, behavior, owner, or scope changes. Clear or narrow stale verification claims. Do not add metadata timestamps.

Profiles and conflicts

Profiles are explicit, scoped extensions: personal, repository, team, or enterprise. Record their source, applicability, additions, and specific overrides in the project's existing guidance. Profiles can define exact folders, naming, owners, tools, and extra non-temporal metadata under extensions.

Combine compatible rules. An applicable explicit override can replace a stated default; neither an organization label nor a more polished document establishes precedence. Use an already declared precedence policy. If two applicable mandatory rules conflict without one, surface both rules and their sources before the dependent edit. Do not silently adopt one. Direct user instructions control the requested scope. Preserve truthful verification, canonical authority, and lifecycle distinctions when extending profiles.

This lightweight schema borrows selected knowledge-management concepts. It does not adopt OKF or claim OKF compliance. Do not import an entire external framework or private organizational taxonomy.

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.