Content
92%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.
An exceptionally well-constructed operational skill: executable verified commands, per-step completion checkpoints, a classification-correction feedback loop, and a closing checklist, with the interpretive detail correctly offloaded to a single real reference file. The only weakness is minor overlap between step 5 and the reading grid on the trailer caveat, which keeps conciseness just off the top anchor.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational — script-specific semantics like "files older than the window take their first commit in the window as creation date" and "exits non-zero without writing" earn their tokens — but step 5's trailer-as-lower-bound caveat restates reading-grid rows, a minor instance that could be trimmed rather than the every-token-earns-its-place top anchor. | 4 / 5 |
Actionability | Every command is copy-paste ready and verified against the actual bundle — "python3 -I ${CLAUDE_SKILL_DIR}/scripts/harness_archaeology.py --repo {repo_path} --out {out} --ref {ref} [...]", the ls-files/grep pipeline for missed files, and the correction flags --extra-glob/--exclude-glob all exist in the script's argparse, and common cases are covered. | 5 / 5 |
Workflow Clarity | Six sequenced steps each end in an explicit "Done when" checkpoint, step 3 is a full validate-then-fix-then-rerun feedback loop for classification errors ("Rerun with --extra-glob ... or --exclude-glob"), and the closing six-item validation checklist covers the whole run; the workflow is read-only, so the destructive/batch cap does not apply. | 5 / 5 |
Progressive Disclosure | The ~55-line body stays an overview-plus-workflow with one clearly signaled, one-level-deep reference — "[the reading grid](references/reading-grid.md)", verified to exist and holding the 12-indicator interpretation table — while the script is properly invoked via ${CLAUDE_SKILL_DIR}/scripts/, making content appropriately split and easy to navigate. | 5 / 5 |
Total | 19 / 20 Passed |