Content
81%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 highly actionable, well-sequenced operational skill body with strong validation gates and a verified one-level-deep reference bundle. Its main weakness is token efficiency: the artifact-resolution mandate and report-format rules are repeated across multiple sections, and the full report template could live in a reference file.
Suggestions
State the artifact-resolution requirement once (in §4.55) and reference it from Non-Negotiables, §3, and Finish Criteria instead of restating the five required fields in each place.
Move the ~75-line §7 report template to a reference file (e.g. references/report-template.md) and keep only the section skeleton and cleanup rules inline.
Add the missing THIRD_PARTY_NOTICES.md to the bundle or remove the instruction to keep it with the skill when packaging.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with Deepline-specific operational knowledge Claude would not know, but the artifact-resolution rule is stated four times (Non-Negotiables, §3's "record why it matters; the exact artifact", §4.55, and Finish Criteria) and the §7 report template runs roughly 75 lines inline. This fits anchor 3 (mostly efficient but some unnecessary repetition that could be tightened) better than anchor 4's "minor instances". | 3 / 5 |
Actionability | Guidance is fully executable throughout: concrete `deepline search "topic -> recent web and news sources" --categories research` examples, complete curl calls to the planning endpoint with headers and payload, a copy-paste SKILL_ROOT resolution loop plus `python3 "$SKILL_ROOT/scripts/query_design.py" "$OBJECTIVE" --depth default`, and `--dry-run` price checks before paid probes. This matches anchor 5 (copy-paste ready commands covering common cases). | 5 / 5 |
Workflow Clarity | The Standard Flow is clearly sequenced (1–7 with explicit gates §4.25, §4.5, §4.55, §4.6), includes validation checkpoints (standalone billing check first, describe-before-pricing, coverage gate over every source family, approval gate with exact question before any full paid run) and an error-recovery feedback loop ("A cell filled from memory instead of a live result is not verified. Search for it."). This matches anchor 5's explicit validation and feedback loops. | 5 / 5 |
Progressive Disclosure | The "Start Here" section clearly signals four real one-level-deep references (source-map.md, last30days-gtm-corpus.md, query-design.md, fanout-consolidation.md, all verified present with substantial content) and three real scripts. Minor gaps keep it at anchor 4 rather than 5: THIRD_PARTY_NOTICES.md is referenced in the Attribution section but absent from the bundle, and the long §7 report template is inlined in SKILL.md rather than split out. | 4 / 5 |
Total | 17 / 20 Passed |