agentleFS
mblaubergmblauberg/provenant/skills/engineering-docs/SKILL.md

engineering-docs

Skill1 starsChanged 4 days agoMITView on GitHub

Use to create, place, index, update or archive engineering docs: specs, ADRs, runbooks, reference, instruction files, kept prototypes and diagrams. Decides durable vs ephemeral and where each belongs. Not for wording (engineering-writing) or session handoffs (session).

2 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

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    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:

    Read the repository process declaration first.

    line 11
  2. Distinctive:

    Keep only what code cannot say, and link code, schemas and tests.

    line 30
  3. Very specific:

    Default to Mermaid; use d2-diagrams for a justified fixed layout.

    line 72
  4. Very specific:

    Check overlap, clipping and lost failure paths, and run mmdc from a temporary directory.

    line 74

≈ “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

Engineering docs

This skill decides what docs are worth keeping and where they live; engineering-writing shapes the prose.

Read the repository process declaration first. If it is absent, use an unambiguous existing owner or propose paths; never invent authority. With an issue-tracker home, the parent issue owns scope and status; docs own durable intent and invariants, and no second status projection is created. A link-only issue map is a fallback for unavailable or cross-tracker routes only.

Durable or ephemeral

KindFate
Issue-body spec, plan, handoff, scratch notesEphemeral. Dies at delivery; its settled rules move into a durable doc in the same PR
Invariants, non-derivable intent, ordering, failure semanticsDurable spec
Architecture: crate and trust boundaries, protocols, storage, major dependenciesLiving ADR
Operator stepsRunbook, linked to its source and verification
Implemented fields, flags, schemas, commandsCode, or generated reference. Never hand-edit it
Other settled decisionsThe relevant doc (spec, runbook, skill). Ephemeral ones stay in the PR or issue thread
What changed and whengit log; release fragments per repository policy

A spec that code can regenerate, or a task spec that has shipped, should die. Keep only what code cannot say, and link code, schemas and tests.

Binding rules

  • Edit in place. A living doc states current truth. Rewrite or delete superseded text in the same change; never append dated Update: or "as of" entries to a spec, runbook, README or ADR. A diff that only adds lines to a living doc needs a reason. Changelogs and release notes are the exception.
  • Owner and anchor. Each spec and runbook opens with Owner: (the responsible crate, skill or document) and Anchor: (relative links to source symbols or tests that let a reader check its claims). Name the symbol, not a line number.
  • Prevent drift. Mechanical owner, anchor, link and generated-diff checks may fail CI; whether prose belongs in a spec is review judgement. Before cutting apparent restatement, check renames and live callers.
  • Three hops from claim to owner to evidence. Merge duplicate truths and tiny files that change together. A doc past about 15 KB prompts review, not a split.
  • Repair indexes and inbound links together. Quote and link frozen legal, compliance or gate wording.

ADRs: architecture only

Record the decision, its forces and consequences, for architecture alone. Process goes to its skill or HARNESS.md; deferrals and product calls to their issues. Amend an accepted ADR in place to the current decision, keeping one Changed: YYYY-MM-DD, #issue line that each amendment replaces. Merge or delete superseded ADRs, fix inbound links, index live decisions only, and leave no compatibility stubs.

Kept prototypes

A prototype kept as reference lives in the repo with a one-line index entry in its owning doc (what it proves, how to open it). It may be shared as a private Claude artifact; the repo file stays the source. Throwaway ones are not indexed.

Instruction files

The root file routes. Add a nested AGENTS.md (10 to 40 lines) only at a real boundary after an observed failure. Each CLAUDE.md holds exactly @AGENTS.md.

Diagrams and archive

Default to Mermaid; use d2-diagrams for a justified fixed layout. Render the exact block and open the output: a clean parse is not visual verification. Check overlap, clipping and lost failure paths, and run mmdc from a temporary directory. Give each block accTitle and accDescr. Archive retired docs with owner authority and repair references; keep a tombstone only for an external link.

See placement for default homes (they apply after setup-repo's approved Declare-the-process step; otherwise propose paths).

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.