Content
100%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-architected instruction skill: lean overview body, concrete executable commands with matching script CLI, a sequenced activation flow with fallback handling, and a clean one-level reference structure. It avoids explaining concepts Claude already knows and keeps all real detail in progressively-disclosed files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense, instruction-led prose that assumes Claude's competence — it never explains what agents/parties are and every section (conventions, activation, party-feel, modes, wrap-up) carries actionable guidance rather than padding. | 3 / 3 |
Actionability | Concrete executable commands throughout — 'uv run {skill-root}/scripts/resolve_party.py --project-root {project-root} --skill {skill-root}', flags '--party <id>', '--list-groups', '--mode <session|auto|subagent|agent-team>', '--non-interactive' — and the referenced script's CLI matches what is documented. | 3 / 3 |
Workflow Clarity | 'On Activation' is a numbered 1–7 sequence with explicit error-recovery checkpoints — 'On failure, read customize.toml directly and use defaults', 'if installed_agents_resolved is false... tell the user, carry on... and improvise' — and a sequenced wrap-up. | 3 / 3 |
Progressive Disclosure | SKILL.md is an overview that loads detailed material on demand via clearly signaled one-level-deep references (create-party, party-memory, mode-auto/subagent/agent-team), all of which exist; modes and memory are appropriately split into separate files. | 3 / 3 |
Total | 12 / 12 Passed |