Content
75%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 well-structured orchestration skill: the workflow is clearly sequenced with explicit gates and user checkpoints, and detail is properly delegated to reference files rather than inlined. The main gaps are the absence of the referenced files from the bundle itself, a few dense multi-clause sentences, and no error-recovery loops for failure paths.
Suggestions
Include the referenced files (01-candidate-profile.md through 09-web-research.md, tools/posting_guard.py, documents/README.md) in the skill bundle so the one-level-deep navigation actually lands, or consolidate paths that live outside the skill directory.
Split the ~120-word Step 3b bullet into short sub-bullets matching the /apply Step 6b fields, and trim maintainer-rationale asides so every line is executing-agent instruction.
Add brief failure-path guidance (what to do when the 09-web-research.md escalation order is exhausted, or when a required CV variant or template is missing) to turn the gates into true validate-and-recover checkpoints.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is operational and delegates detail to reference files rather than teaching known concepts, but a few passages could be tightened: the Step 3b bullet is a single ~120-word sentence, and asides like 'This step exists here because /scrape Step 5 routes straight into this skill' are maintainer rationale rather than executing-agent instruction. | 4 / 5 |
Actionability | Concrete guidance is present (exact output paths like 'cv/main_<company>_<role>.tex', a runnable command with flags 'python3 tools/posting_guard.py' with --html/--text, and explicit stop and verbatim-retention rules), but the templates and scoring frameworks the steps depend on live in referenced files absent from the bundle, and no example invocation output (e.g., the verdict line format) is shown inline. | 4 / 5 |
Workflow Clarity | The sequence (Step 1, 2, 3, 3b, 4) is clearly ordered with explicit gates ('Gate it before reading it as content', 'Run this once both documents exist', 'stop before creating any file') and an ask-user checkpoint, but there is no error-recovery feedback loop for failure paths such as the fetch escalation order exhausting or a referenced template being missing. | 4 / 5 |
Progressive Disclosure | The Reference Files table gives well-signaled, one-level-deep references with per-file purposes, but the referenced files (01-09) are not present in the bundle being evaluated, and several referenced paths (tools/, documents/, .claude/commands/apply.md) fall outside the skill's own directory, so navigation cannot be fully verified. | 4 / 5 |
Total | 16 / 20 Passed |