Content
88%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 lean, command-dense body with a well-gated workflow and real failure-recovery loops for tests and installs. The one material weakness is that its most-referenced file, reference-pytest.md, is absent from the skill bundle, leaving the per-package pytest command matrix unreachable from the provided files.
Suggestions
Ship reference-pytest.md in the bundle (e.g. references/reference-pytest.md) containing the promised sections '#1-re-run-with-visible-logs' and '#when-tests-fail', or inline the per-package pytest cwd/PYTHONPATH commands.
Make the link resolve unambiguously relative to SKILL.md (e.g. [reference-pytest.md](references/reference-pytest.md)) so the anchors cited throughout the body actually work.
In the Layout section, note where reference-pytest.md lives so a reader can locate it from the file listing even if a link breaks.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and imperative throughout — e.g. 'Installed tentacles output: repo-root tentacles/ (generated — never edit)' — with zero explanation of concepts Claude already knows; conventions are delegated to CONTRIBUTING-agent.md and AGENTS.md instead of restated. Matches the 'lean and efficient; every token earns its place' anchor; there is no over-explanation to trim that would justify a 4. | 5 / 5 |
Actionability | Concrete copy-paste commands cover the common cases: 'source .cursor/env.sh', 'bash .cursor/reinstall-tentacles.sh', 'python -m tools.extended_linter --base origin/dev', 'bash .cursor/cloud-validate.sh --profile ci-tentacles --json', 'bash .cursor/seed-agent.sh --full'. Not 5 because the most common operation is deferred — 'Run targeted pytest per reference-pytest.md (CI matrix)' — and reference-pytest.md is not present in the skill bundle, so the actual pytest invocations are unavailable; not 3 because everything else is fully executable. | 4 / 5 |
Workflow Clarity | A numbered 7-step workflow with explicit validation gates — '**Success:** Build log contains OCTOBOT_CLOUD_INSTALL_OK. Do not guess from UI alone', 'Before handoff: python -m tools.extended_linter --base origin/dev' — and a genuine error-recovery loop ('**If any test fails,** start with step 1 — visible logs', 'do not widen mocks, seeds, or timeouts until that boundary is clear'). Matches the anchor for clear sequence with explicit validation and feedback loops. | 5 / 5 |
Progressive Disclosure | Good structure with clearly signaled, one-level-deep references, each annotated with what to find there: '**Python (imports, literals, public API):** [CONTRIBUTING-agent.md]', 'Machine rules: tools/extended_linter/config/policy.yaml', per-package AGENTS.md pointers. Not 5 because the bundle contains no references/ directory — reference-pytest.md, cited about six times with specific anchors ('#when-tests-fail', '#1-re-run-with-visible-logs'), is not a real file in the provided bundle, so the primary navigation target is unverifiable; not 3 because the references that are signaled are clear and well placed. | 4 / 5 |
Total | 18 / 20 Passed |