Reviews, designs, and improves Claude Code skills. Use when someone wants feedback on an existing skill, needs help designing a new skill's architecture and file structure, or wants to define a testing strategy for a skill. Also trigger when users say things like "review my skill," "how should I structure this skill," "what should I test," "skill assessment," "skill architecture," or "evaluate this skill."
Multi-perspective review and design for Claude Code skills. Five assessor subagents independently analyze a skill or skill proposal through distinct lenses, then a lead synthesizer produces a unified recommendation.
Three task modes:
If the user's request doesn't clearly map to one mode, ask which they'd like — or suggest the one that seems most relevant based on context.
Each assessor is a distinct persona with a primary lens and secondary awareness of other concerns. They are not single-axis evaluators — they're modeled after real team members who lead with one perspective but notice issues across the board.
Each assessor runs as a subagent defined in agents/assessor.md. Each receives their persona file from agents/personas/ along with the shared task inputs.
| Persona | File | Focus |
|---|---|---|
| Riley — The Architect | agents/personas/riley.md | Structural soundness, progressive disclosure, file organization |
| Morgan — The Consumer Advocate | agents/personas/morgan.md | Agent consumability, context efficiency, routing clarity |
| Jordan — The Steward | agents/personas/jordan.md | Long-term skill health: maintainability, testability, failure modes, naming, detail preservation |
| Alex — The User Advocate | agents/personas/alex.md | End-user experience: interaction flow, output usefulness, error recovery, first-time vs. repeat use |
| Avery — The Strategist | agents/personas/avery.md | Scope, triggering accuracy, strategic fit, ecosystem positioning |
| Persona | File | Focus |
|---|---|---|
| Sage — The Creative Provocateur | agents/personas/sage.md | Novel approaches, unconventional framings, cross-domain analogies |
Activate Sage when the user says "add Sage," "I want the creative take," "give me the wild ideas," "+1," or when you sense the problem space would benefit from a fundamentally different angle — especially for Architecture mode.
When Sage is active, spawn 6 subagents instead of 5. The synthesizer gives Sage's novel proposals their own section in the output — "Creative Alternatives" with tradeoff analysis, not blended into the convergence analysis.
Determine which mode the user needs: feedback, architecture, or testing. If ambiguous, ask.
Depending on mode:
If the total input materials from Step 2 exceed ~500 lines, produce a structured briefing before spawning assessors:
Assessors receive the briefing as their primary input, with pointers to the original files for targeted deep-reads when they need specific details.
If input materials are under ~500 lines, skip this step and pass the original files directly to assessors.
Spawn the assessor panel in parallel:
Each subagent is spawned using the agents/assessor.md agent definition. Each receives:
agents/personas/{name}.md)agents/task-briefs/{mode}.md)Assessors load their own reference materials (evaluation criteria, skill design principles, advanced progressive disclosure) as part of their assessment process. All subagents receive the same task description (mode + user context + what to assess). The differentiation comes from each persona's thinking style and rubric dimension assignments.
Once all assessors complete, spawn the synthesizer using agents/synthesizer.md. The synthesizer receives:
The synthesizer produces the final deliverable as structured JSON, which is then used to generate an HTML report.
After synthesis, generate a styled HTML report:
scripts/generate-report.pyreport-data.jsonpython scripts/generate-report.py report-data.json --openThe report format varies by mode:
| Mode | Output Format |
|---|---|
| Feedback | Structured markdown report with convergence/divergence analysis, prioritized surgical recommendations, and a rubric scorecard. If structural issues are identified, leads with a restructure proposal. |
| Architecture | Interactive walkthrough — present the proposed structure, explain key decisions, invite discussion before finalizing |
| Testing | Testing strategy with scored rubric, coverage analysis, metrics framework, and phased testing plan. Saves testing-strategy.md for skill-creator handoff. |
After presenting, invite the user to discuss, ask questions, and iterate on the proposal.
After the user has worked through discussion and made decisions, automatically consolidate what was agreed upon:
Then offer the user a choice:
project-plan.md with implementation steps sequenced by dependency and priority.What "implement now" means per mode:
| Mode | Consolidated output | What gets built |
|---|---|---|
| Architecture | architecture-spec.md — finalized structure, file layout, output contract | Draft the complete skill: SKILL.md + all reference files. Spawn subagents to write files in parallel where possible. |
| Feedback | feedback-recommendations.md — prioritized changes with specific edit locations | Apply the recommended changes to the skill files. Work through P0 items first, then P1. |
| Testing | testing-strategy.md — coverage priorities, metrics, flakiness expectations, skill gaps | Save testing-strategy.md to the skill directory. Offer handoff to skill-creator for eval writing. |
If subagent spawning fails or the environment doesn't support it:
This is less rigorous (later assessors have context of earlier ones, reducing independence), but still provides multi-perspective value.
testing-strategy.md file to the skill's directory; the skill-creator checks for this file when writing evals and uses it to guide eval design (coverage priorities, metrics, flakiness expectations). If no strategy file exists, the skill-creator writes evals from scratch.| Path | Contents |
|---|---|
agents/assessor.md | Shared assessor agent definition (process, output format, guidelines) |
agents/synthesizer.md | Lead synthesizer agent definition |
agents/personas/ | Individual persona files (riley, morgan, jordan, alex, avery, sage) |
agents/task-briefs/ | Mode-specific task briefs (architecture, feedback, testing) |
references/ | Evaluation criteria, skill design principles, advanced progressive disclosure |
references/techniques/ | Reusable improvement techniques with a symptom-based routing index |
scripts/generate-report.py | Generates styled HTML report from synthesis JSON output |
assets/report-template.html | HTML template for the report |