Content
56%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 highly actionable in shape — exact commands, documented flags, defined JSON fields, and a full output template — and the workflow sequence is clear with troubleshooting recovery paths. Its weaknesses are verbosity (network adaptation covered three times, source lists and keyword taxonomies inlined that duplicate the bundle's scripts and references) and a path inconsistency ('technology-news-search' vs. name 'technology-search') that would break the documented commands as written.
Suggestions
Deduplicate the network-adaptation material into a single section and drop the verbatim 18-source list (it appears twice); state only the behavior — global network → all 75 sources, China-only → 18 China sources, 3s detection with 5-minute cache.
Move the domain keyword/alias taxonomy and heat-score formula out of SKILL.md into references (e.g., references/domains.md) or rely on the existing scripts/shared/domain_classifier.js and heat_calculator.js, keeping only a few illustrative routing examples inline.
Fix the script path inconsistency: the skill's name field is 'technology-search' but every command uses "$SKILLS_ROOT/technology-news-search/scripts/search-news.sh" — align one of them so the documented invocations are executable as written.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably padded: network adaptation is explained three separate times (Overview, Smart Source Routing, and a third Network Adaptation block), the identical 18 China-source list is enumerated twice verbatim, and the ~150-keyword domain taxonomy and heat-score formula duplicate logic that lives in scripts/shared/domain_classifier.js and scripts/shared/heat_calculator.js. It does not explain concepts Claude already knows, so it stays above anchor 1, but the repeated and mis-filed material matches 'noticeably verbose; several unnecessary explanations or padded sections'. | 2 / 5 |
Actionability | Commands are concrete and copy-paste shaped (bash "$SKILLS_ROOT/technology-news-search/scripts/search-news.sh" "Electron" --limit 15 with documented --limit/--max-per-source/--no-balance/--all-sources flags), but there is a real execution gap: every command path uses 'technology-news-search' while the skill's name field is 'technology-search', so the documented invocations would not resolve as written. Missing key executable details like this places it at 'some concrete guidance but incomplete' rather than anchor 4. | 3 / 5 |
Workflow Clarity | The Workflow section gives a clear numbered sequence (extract keyword → run script → read JSON output → translate → group by heat tier → present Markdown) and a Troubleshooting section provides error-recovery paths for empty results, script errors, and slowness. It is a read-only search (no destructive/batch cap applies), but checkpoints like 'verify results are non-empty before formatting' are only implicit in troubleshooting, so it fits anchor 4 rather than 5. | 4 / 5 |
Progressive Disclosure | The body does signal real one-level-deep references ([references/sources.json](references/sources.json), which exists, and the scripts/ tree, which matches the bundle listing). However, content that clearly belongs in the bundle is inlined in SKILL.md: the full 75-source roster, the entire domain keyword/alias taxonomy, and the detailed heat-score formula. That mix of clear signaling with substantial misplaced inline content matches anchor 3 rather than anchor 4's 'most content is appropriately placed'. | 3 / 5 |
Total | 12 / 20 Passed |