Content
93%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A concise, actionable body that defers flag detail to the bundled script and `--help` while surfacing non-obvious conventions and safety rails. The only gap is the absence of an explicit error-recovery loop on the destructive remove path.
Suggestions
Add a short validate→fix→retry note for `remove` (e.g. on a dirty refusal: commit/stash or inspect, then re-run) to make the recovery path explicit and lift workflow_clarity to 5.
Optionally show one line of expected `create` final-line output so callers know the absolute path is the cd target without reading the script.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Lean ~35-line body that assumes Claude's git knowledge; every line covers skill-specific conventions (sandbox-writable roots, .git/info/exclude auto-exclusion, commits survive teardown) with no general-concept padding. | 5 / 5 |
Actionability | Provides copy-paste-ready commands (`uv run .../worktree.py create|list|remove [flags]`) covering the common cases, plus a pointer to `--help` for exact flags. | 5 / 5 |
Workflow Clarity | The destructive `remove` path documents validation gates (refuses if dirty, refuses if unmerged, 'Never force without first investigating what would be lost'), so it is not capped at 3; however no explicit validate→fix→retry feedback loop is described, landing at anchor 4 rather than 5. | 4 / 5 |
Progressive Disclosure | Well-organized into Run it / Conventions / Gotchas / Resources with a single, clearly signaled one-level-deep reference to the bundled `scripts/worktree.py` ('run it, don't read it'; use `--help`), keeping detail out of the body. | 5 / 5 |
Total | 19 / 20 Passed |