Content
86%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 skill body: executable commands for every invocation mode, a clearly sequenced three-phase workflow with data-quality checkpoints, and textbook progressive disclosure into two real, one-level-deep reference files. The main improvements are removing verbatim caveat repetition across Overview/Do-NOT-use/Guardrails and adding an explicit output-verification step before presenting the report.
Suggestions
Consolidate the repeated 'crowding is not a trade signal' warning: state it once in Guardrails and reference it from the Overview/Do-NOT-use sections with a pointer instead of restating it three more times — same for the 'fade speculators, not commercials' rationale duplicated between Overview and Guardrails.
Add an explicit verification checkpoint at the end of Phase 1 (e.g., 'Confirm reports/cot_crowding_<date>.json and .md exist and the run_context.data_date matches the latest Friday release; check the skipped list before presenting') to close the workflow_clarity gap.
In Phase 3, replace the pointer to a separate 'position-sizer' skill with either a concrete inline sizing rule (e.g., fixed fractional risk per trade) or a note that the skill may not be installed, so the guidance stays actionable standalone.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean — methodology specifics (Shapiro's five steps, the 3-day publication lag, the 156w/26w windows) are domain knowledge Claude does not already have and earn their tokens, and code/command blocks are minimal. It sits at 4 rather than 5 because key caveats are repeated nearly verbatim across sections: 'crowding is not a trade signal' appears in the Overview, 'Do NOT use when', Phase 3, and Guardrails, and 'fade speculators, not commercials' appears in both the Overview and Guardrails — one of each could be trimmed. | 4 / 5 |
Actionability | Phase 1 gives copy-paste-ready commands covering the common cases ("--core", "--symbols \"ES,GC,CL\"", and full-universe invocations with --output-dir), prerequisites name the exact env var and plan tier, and the Output section specifies the exact artifact paths and their contents. This matches the 'fully executable; copy-paste ready; covers common cases' anchor. | 5 / 5 |
Workflow Clarity | Phases 1-3 are clearly sequenced and include real checkpoints: insufficient-history markets surface in a 'skipped' list with reasons rather than being silently dropped, the Cadence section flags the 3-9 day data lag, and the guardrail gates any entry recommendation on steps 2-3 confirming. It falls short of the 5 anchor because there is no explicit verify-the-output step before presenting (e.g., confirm the report file was generated and the as-of date is the expected Friday release). | 4 / 5 |
Progressive Disclosure | SKILL.md is a genuine overview with details split into two references (shapiro-methodology.md, cot-index-calculation.md), both verified to exist and only one level deep; each is described with its contents and when to load it, and the 'When to Load References' section makes navigation explicit while noting regular execution needs none. This matches the 5 anchor's clear overview with well-signaled one-level-deep references. | 5 / 5 |
Total | 18 / 20 Passed |