Content
67%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 dense, well-structured operations manual with genuinely concrete thresholds, dispatch rules, and output-format specifications, and it never wastes tokens explaining concepts Claude already knows. Its weaknesses are duplicated framing content (the HTML comment block and overlapping trigger/keyword tables) and a progressive-disclosure chain whose referenced files are absent from the bundle, which together keep it just below the top tier.
Suggestions
Delete or drastically de-duplicate the HTML comment block (CAPABILITIES_SUMMARY / COLLABORATION_PATTERNS / BIDIRECTIONAL_PARTNERS / PROJECT_AFFINITY) — it restates the body's Collaboration and Trigger Guidance sections at a cost of roughly 40 lines.
Merge the 18-bullet Trigger Guidance list with the Output Routing keyword-cue table, which covers nearly the same signals in tabular form.
Ship the referenced reference/ and _common/ files with the skill bundle (or prune the pointers to only files that exist) so the progressive-disclosure chain resolves instead of dead-ending at every "Read" column entry.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The prose is terse and assumes Claude's competence (SWOT, TAM, OSINT, and SCIP are never explained), but the ~43-line HTML comment block (CAPABILITIES_SUMMARY, COLLABORATION_PATTERNS, BIDIRECTIONAL_PARTNERS, PROJECT_AFFINITY) restates the body's Collaboration and Trigger Guidance sections, and the 18-bullet Trigger Guidance list overlaps the Output Routing keyword-cue table. These duplications are more than the "minor instances that could be trimmed" of anchor 4. | 3 / 5 |
Actionability | For an instruction-only skill the guidance is strongly concrete: "Define 5-10 Key Intelligence Questions", "Actively track 3-5 primary competitors; passively monitor 10-15 via automated alerts", "10%+ price cut = High", and an exact citation format example "[1] https://example.com/pricing — accessed 2026-03-27". It falls short of fully executable because key mechanics (per-Recipe behavior, engine prompts, most numeric thresholds) defer to reference files that are not present in the bundle. | 4 / 5 |
Workflow Clarity | Multiple workflows are clearly sequenced as phase tables with required actions and key rules (MAP → ANALYZE → DIFFERENTIATE; DISCOVER → POSITION → CRAFT → AMPLIFY → MEASURE), with ordering constraints ("KIQs before collection; WebSearch first, then source list before analysis") and a SHARPEN calibration feedback loop. Not 5 because checkpoints are expressed as rules rather than explicit validate → fix → retry steps, and several parallel workflow definitions require the reader to disambiguate via the Output Routing table. | 4 / 5 |
Progressive Disclosure | Disclosure is well signaled and one level deep: per-shape "Read" columns, a Reference Map with read-triggers, a recipes index, and a full-index pointer, with the body kept as an overview. It cannot score 5 because none of the referenced files (reference/*.md, _common/*.md) ship in the bundle, so the navigation chain cannot actually be followed; it stays above 3 because the signaling and overview-level organization themselves are strong. | 4 / 5 |
Total | 15 / 20 Passed |