git:20260804.13d3632 to git:20260828.23e381d

4 added, 0 removed. Audit A to A.

# CLAUDE.md

+ ## SKILL.md placement rule

+ New behavior docs go in `plugins/gh-review-loop/skills/gh-review-loop/references/<topic>.md`; SKILL.md gets at most one pointer line. Inline in SKILL.md only what the agent needs every cycle. Size budgets are enforced by `tests/test_doc_budgets.py` against `tests/budgets.json` — raising a budget must happen in the same PR, with justification.

## Test runner

```bash

# If installed via Homebrew on Apple Silicon:

/opt/homebrew/bin/pytest

# Or run pytest from a virtual environment:

pytest

```

`python3 -m pytest` does not work — pytest is not on the bare Python path.

## Python type annotations

All scripts use `from __future__ import annotations`. This makes subscripted generics (`dict[str, Any]`, `list[str]`, `str | None`) safe on Python 3.9+. Do not revert to bare `dict` / `list` to "fix" compatibility — it is already handled.

## Plugin cache

The installed plugin lives at:

```

~/.claude/plugins/cache/gh-review-loop/gh-review-loop/<version>/

```

To test unreleased script changes without merging, copy the modified files directly into the cache. A future `/plugin` update will overwrite them.

## Scripts location

```

plugins/gh-review-loop/skills/gh-review-loop/scripts/

```

When running scripts locally (not through the installed plugin), invoke from the repo root:

```bash

python3 plugins/gh-review-loop/skills/gh-review-loop/scripts/fetch_gemini_threads.py

```