agentsclimarketplace

New stage

Skill tanker327/agent-skill-analysis/.claude/skills/new-stage

Scaffold a pipeline stage for the SkillAnalysis analyzer — src module, mirrored test file, diagnostic registry entries, fixtures, and wiring into analyze.tsFrom its SKILL.md

Install
npx -y skills add tanker327/agent-skill-analysis --skill new-stage

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.9 KB, 662 tokens by cl100k_base, as published. Nobody here has run it

Scaffold a new pipeline stage named $1 (kebab-case → src/$1.ts), following the conventions from global_ignore/skl-skill-analysis-plan.md §2–§3 and .claude/rules/testing.md. Any further arguments are diagnostic codes the stage will emit.

Preconditions — verify before doing anything

  1. src/schema.ts, src/diagnostics.ts, and src/analyze.ts exist (P0 has landed). If not, STOP and tell the user this skill needs the P0 skeleton first.
  2. src/$1.ts does not already exist. If it does, STOP — this skill scaffolds, it does not overwrite.

Steps

  1. Create src/$1.ts — a pure function module:

    • One exported function taking plain inputs (never a SkillSource — IO stays in analyze.ts) and returning { <result fields>, diagnostics: Diagnostic[] }.
    • No clock, no randomness, no locale-dependent calls, no node: imports.
    • Any array it returns must have a fixed, documented sort order.
    • Match the style of the closest existing stage module.
  2. Register diagnostic codes — for each code passed as an argument, add an entry to the registry in src/diagnostics.ts: code, default severity (error | warning), message template, optional hint. Codes are kebab-case, no E_/W_ prefixes (e.g. readme-missing).

  3. Create tests/$1.test.ts — mirrored test file:

    • Table-driven unit tests over the pure function: happy path, empty input, and one case per diagnostic code (the vocabulary meta-test in tests/contract.test.ts fails CI if a registered code has no fixture).
    • Zero mocks. Use plain values or the fromFiles memory source via tests/helpers.ts for full-pipeline cases.
  4. Wire into the pipeline — call the stage from src/analyze.ts at the right point in the flow (see the pipeline order in global_ignore/skl-skill-analysis-flow.md §1, or the stage sequence already in analyze.ts). If the stage adds output fields, update SkillAnalysisSchema in src/schema.ts in this same change and extend the full-output fixtures.

  5. Run the gatenpm run lint && npm run typecheck && npm test. All green, including the contract meta-tests (determinism, schema conformance, diagnostic coverage).

  6. Report — list created/modified files, the registered codes, and remind the user this should be committed per .claude/rules/commit-message.md (scope = the stage name; arch type if the schema changed, otherwise feat).

Stay within the current build-plan phase: if $1 belongs to a later phase than the one in progress, point that out and ask before proceeding.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 326,736. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.