Content
82%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-engineered body: fully executable commands, a gated multi-step workflow, and clearly signaled one-level-deep references. The main costs are duplicated boundary-keeping (Do-NOT-use vs the Distinct-from table), an inline forcing-question library that should be a reference file, and a missing explicit recovery step for failed triangulation.
Suggestions
Move the forcing-question library (lines 122–146) to a reference file such as references/forcing_questions.md and keep a one-line pointer in SKILL.md — it trims ~25 lines from the always-loaded context and improves progressive disclosure.
Merge the "Do NOT use this skill to" paragraph with the "Distinct from" table into a single boundary section to remove the duplicated neighbor-skill enumeration.
Add an explicit recovery step to the sizing workflow for a failed triangulation flag (e.g., revisit assumptions register → adjust serviceable fraction → re-run market_sizer.py → re-check divergence) to close the validate-fix-retry loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and skill-specific — no space is spent explaining concepts Claude already knows (Kotler, Cochran, TAM are cited, not taught). It falls short of lean because the "Do NOT use this skill to…" paragraph duplicates the "Distinct from" table, and the ~25-line forcing-question library is inline content that could live in a reference file. | 4 / 5 |
Actionability | Commands are copy-paste ready and cover the common cases: `market_sizer.py --input market.json --method both --profile {b2b-saas|consumer|…}`, `sample_size_planner.py --population 62000 --confidence 0.95 --moe 0.05`, `python3 scripts/onboard.py --show`, a complete `/ar:setup` block, and a documented `--help`/`--sample`/`--output {human,json}` contract for every tool. | 5 / 5 |
Workflow Clarity | The 5-step workflow is clearly sequenced with substantive checkpoints — "Reconcile the top-down/bottoms-up delta before quoting anything", "Fund the per-segment floors", "Drop segments failing the substantiality/accessibility gate" — and the sizer "flags failed triangulation". It stops short of 5 because the main workflow lacks an explicit recovery loop (what to do after a failed triangulation flag other than "reconcile"), and the brief step has no validation before sizing proceeds. | 4 / 5 |
Progressive Disclosure | Structure is good: a dedicated References section lists three clearly-labeled one-level-deep files, scripts are tabulated, and the asset template is named. However, no bundle files (references/, scripts/, assets/) are present in this copy to verify, and the inline forcing-question library is content that belongs in a separate reference file, which keeps it below the clear-overview 5 anchor. | 4 / 5 |
Total | 17 / 20 Passed |