← ClaudeAtlas

iatlisted

Audit Python scripts for compliance with the Inline Audit Trail (IAT) convention — INTENT, REASONING, ASSUMES, SOURCE comments and validation guards — AND propose the comment text to add. LLM-driven, judgment-heavy, slower than /check. Use when the user invokes /iat, /iat <script>, /iat <slug>, or asks to 'check IAT compliance', 'audit comments in source/', 'add missing IAT comments'.
hsigstad/research-kit · ★ 0 · AI & Automation · score 75
Install: claude install-skill hsigstad/research-kit
# /iat — Inline Audit Trail compliance check (with proposed comments) Audits data-transformation scripts against the Inline Audit Trail convention defined in `research-kit/rules/inline_audit_trail.md`, and proposes the exact comment text to add for every weak spot. Judgment-heavy and slower than `/check` — that's why it lives as its own skill. The default behavior is *propose, don't edit*. The skill emits the suggested comment text with the line number; the researcher pastes it in. IAT comments are subtle and benefit from a human read before they land. ## How to invoke | Form | What it does | |------|--------------| | `/iat` | Audit all `source/**.py` in the current project | | `/iat <script>` | Audit one script | | `/iat <slug>` | Audit a specific project by slug | | `/iat --no-suggestions` | Report only — no proposed comment text | ## Scope By default, audit all `.py` files in `source/`. IAT applies to **data transformation scripts** only. Skip: - `diarios/` module code (shared library, separate convention). - Configuration files, SCons build files. - LaTeX sources. - Notebook cells unless they contain substantial pipeline code. - Pure utility/helper modules that don't do data work. - Underscore-prefixed scripts (`_foo_check.py`) — exploratory; opted out. ## What to check For each script, evaluate these six categories. Be **conservative**: only flag what would genuinely surprise a careful reader who didn't write the code. Over-commenting is also a problem. ### 1.