Content
68%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-organized advisory skill body with executable, verified tooling and strong reference signaling via the when-to-load table. Its weaknesses are the absence of an explicit multi-step workflow with validation checkpoints and a few trim/split opportunities, plus one reference that cannot be resolved inside this bundle.
Suggestions
Add an explicit end-to-end workflow with checkpoints for the main tasks, e.g. PMF assessment: 1. read company-context.md if present, 2. run `python scripts/pmf_scorer.py --input data.json`, 3. map scores to Invest/Maintain/Kill postures, 4. validate against the Red Flags list before recommending (workflow_clarity).
Fix or inline the dangling `../agent-protocol/SKILL.md` reference — the Internal Quality Loop steps are load-bearing for output quality, so either summarize them in this file or ship the protocol as a bundled reference (progressive_disclosure).
Trim the Keywords section (it duplicates the frontmatter description's triggers) and cut or concretize the generic 'Reasoning Technique: First Principles' section (conciseness).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense, table-driven body that assumes Claude's competence (no basic PMF or north-star explanations), with only minor trim candidates: the Keywords section duplicates the frontmatter triggers and the generic 'Reasoning Technique: First Principles' section adds little skill-specific value. Not a 5 because those few padded sections could be cut; not a 3 since almost everything else earns its tokens. | 4 / 5 |
Actionability | Two verified, copy-paste-ready commands (`python scripts/pmf_scorer.py`, `python scripts/portfolio_analyzer.py`) that run on built-in sample data, plus concrete thresholds ("D30 retention is below 20% (consumer) or 40% (B2B)") and a request→artifact table. Minor gaps keep it below 5: no inline example of the scripts' JSON input format and no concrete procedure for non-script tasks like board prep or org design. | 4 / 5 |
Workflow Clarity | A rough sequence exists ("Always read company-context.md before responding", Quick Start commands, Output Artifacts mapping) but no explicit end-to-end workflow (e.g., collect data → run scorer → map to invest/maintain/kill → cross-check Red Flags) and validation checkpoints are only implicit or deferred to the external agent-protocol file. This matches anchor 3's 'sequence present but checkpoints missing or implicit'; nothing is destructive or batch, so no cap applies, but it also lacks the explicit validation steps of anchor 4. | 3 / 5 |
Progressive Disclosure | Good structure against the actual bundle: a Resources table with a 'When to load' column signals all five existing bundle files (verified present), references are one level deep with no cross-nesting, and detail-heavy material (frameworks, playbooks) is correctly split out. Minor gaps: `../agent-protocol/SKILL.md` dangles outside this bundle, and the ~200-line body inlines sizable dashboard and C-suite-integration tables that could partly live in references. | 4 / 5 |
Total | 15 / 20 Passed |