Content
48%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 cleanly sectioned, safety-conscious overview with an explicit fail-fast workflow, but it stays at the altitude of process maxims: no concrete commands or executable examples anywhere, some duplicated/abstract filler, and a progressive-disclosure pointer to a path that does not exist while the actual bundle reference files go unmentioned. It reads more like a policy document than an operational skill.
Suggestions
Add one concrete worked example to the Workflow or Examples section — e.g., the actual commands for a representative bootstrap (git clone <url> into a worktree, detect and run the repo's install command, run its test command, report the first failing gate) — so the guidance is executable rather than aspirational.
Fix the Progressive Disclosure section to reference the real bundle paths (references/contract.yaml, references/evals.yaml, references/task-profile.json) or move those files to the directory the body names, so deferred content is actually reachable.
Remove the verbatim duplication of the frontmatter description at the top of the body, and either drop or merge the overlapping Philosophy/Avoid/Anti-Patterns bullets to tighten token usage.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The bullet style is terse, but the body repeats the frontmatter description verbatim ("Create, diagnose, and validate a local dev bootstrap. Use when the user asks to..."), and the abstract "Philosophy" section, the cryptic "Schema-bound outputs include `schema_version`" line, and overlapping Avoid/Anti-Patterns bullets add tokens that could be tightened. Not 4: these are more than minor instances — one is pure duplication; not 2: the body never explains concepts Claude already knows and is not padded with prose. | 3 / 5 |
Actionability | Every step is a high-level hint — "Inspect 2-3 focused surfaces", "Take the smallest action that advances the confirmed goal", "Run the narrowest real validator or command path available" — with no concrete commands or examples (e.g., git clone/npm install invocation patterns, sample validation commands). Not 3: even for an instruction-only skill, nothing here is directly executable; not 1 because Inputs/Outputs and the fail-fast rule do give some concrete structure. | 2 / 5 |
Workflow Clarity | The five-step sequence is ordered with explicit checkpoints — "Stop at the first failed gate or blocker and report exact evidence" and "Rerun the relevant validation after fixes before claiming completion" — which is a stated validate→fix→rerun feedback loop, and the separate Validation section reinforces it. Not 5: the steps and checkpoints are abstract rather than command-level, so there is no concrete recovery or checklist detail; not 3: checkpoints are explicit, not missing or implicit. | 4 / 5 |
Progressive Disclosure | The body is a well-sectioned 72-line overview with a dedicated Progressive Disclosure section and a load-only-what's-needed rule, but scored against the actual bundle it stumbles: the referenced path "Infrastructure/references/deferred-skill-context/agent-ops-bootstrap/" does not exist in this bundle, and the real reference files (references/contract.yaml, references/evals.yaml, references/task-profile.json) are never linked from the body. Not 4: navigation to the deferred material is broken/unsignaled; not 2: the SKILL.md itself keeps detail out appropriately and is clearly organized, not monolithic. | 3 / 5 |
Total | 12 / 20 Passed |