Auditar

Skill adriannoes/awesome-agentic-ai/cursor-claude-codex/skills/igoruehara-spec-driven/skills/auditar

329 agent skills (Cursor, Claude Code & Codex), 5,380 OpenClaw skills, 201 ML notebooks, 7 textbooks, 52 research papers, 17 industry reports for PMs, Designers & Developers.

Install

npx -y skills add adriannoes/awesome-agentic-ai --skill auditar

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

What its author says it does

Copied from the file, not written here

Use to audit the conformance of the SDD pipeline — runs the structural validator (frontmatter, links, specs) and checks what requires judgment (AC→test→commit traceability, DoD, orphan specs, live docs up to date). Reports violations with the file. Trigger with /auditar.

SKILL.md

2.2 KB, 479 tokens by cl100k_base, as published. Nobody here has run it

Translation note: Originally authored in Portuguese (pt-BR) by Igor Uehara (igoruehara/spec-driven, MIT). Translated to English by this hub to keep the repository language consistent. Original content unchanged in meaning; see the upstream repo for the pt-BR source.

Skill: Audit the SDD pipeline

Checks whether the project follows the pipeline standard. Two layers: structural (script, deterministic) and semantic (agent judgment).

1. Structural check (deterministic)

Run the validators and report the output:

node scripts/audit-esteira.mjs .
node scripts/validate-mermaid.mjs .

The first covers: frontmatter present + correct dialect (alwaysApply in docs; name+description in skills), broken relative links, and every specs/NNNN-*/ with a spec.md. The second validates the Mermaid blocks (type, quotes, delimiters). Exit ≠ 0 on either = failure.

2. Semantic check (judgment)

The script does not catch everything. Also verify:

Traceability: each AC-N in the spec appears in tasks.md ("Covers AC" column) and has a test?
Orphan specs: features in specs/ with no corresponding PR/implementation, or stalled for a while (cross-check with STATE.md).
Live docs: do the glossary and context-map.md reflect the current terms/boundaries? Do ADRs cover the hard-to-reverse decisions already made?
Pending DoD: features marked as done with an open SPEC_DEVIATION or an AC without a test.
alwaysApply frontmatter: what is foundational is true; the rest false with a description that says when to pull it.

Output

List the violations by file, separating structural (fix now) from semantic (review). Offer to fix the trivial ones (frontmatter, links). Do not invent conformance — report what you found.

This skill is the human/agent counterpart of the CI gate (/setup-ci runs the same script).

What ships with it

Read from the repository

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