Content
65%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.
The body is a maximally lean two-line stub that delegates everything to an external shared skill file via a single absolute path. As a pointer it is perfectly token-efficient and unambiguous, but it provides no actionable audit guidance, no workflow, and its only reference targets a path outside the skill bundle that cannot be verified or shipped with it.
Suggestions
Add a brief overview section (a few lines summarizing the audit dimensions and the expected output) so the skill is not a pure external pointer.
Reference a bundle-local copy of the shared instructions (e.g. references/audit.md) instead of, or in addition to, the absolute ~/.agents path so the skill works standalone.
Include minimal invocation examples for `$optimize-harness [harness ...] [focus]` (e.g. `$optimize-harness claude hooks`) so the delegation is actionable even before the external file is loaded.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is two lines with zero padding: "Load and follow the shared skill at `~/.agents/skills/optimize-harness/SKILL.md`." Every token earns its place and the skill fully assumes Claude's competence. It is not 4 because there is no over-explanation anywhere to trim. | 5 / 5 |
Actionability | The single directive is concrete and executable (an exact file path to load), but all task-level guidance — audit steps, focus-parameter handling, harness specifics — lives outside this file. It is not 4 because the gap is not minor (the file contains zero procedure), and not 2 because the instruction is a specific executable step rather than a high-level hint. | 3 / 5 |
Workflow Clarity | The sequence reduces to "load the shared skill, then follow it", which is unambiguous at the stub level, but the actual audit workflow and any validation checkpoints are entirely delegated and absent from this file. It is not 2 because what is written is coherent and precise, with no poorly defined steps. | 3 / 5 |
Progressive Disclosure | The sole reference is clearly signaled and only one level deep, but the body contains no overview content and points to `~/.agents/skills/optimize-harness/SKILL.md` outside the bundle (no local references/, scripts/, or assets/ exist; the comment's `scripts/agent-harnesses.py` is also not in the bundle). It is not 4 because the entire content split lives outside the skill, so navigation cannot be satisfied from the bundle itself. | 3 / 5 |
Total | 14 / 20 Passed |