Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-architected index skill: strongly actionable, with clearly sequenced workflows and validation checkpoints, and excellent progressive disclosure through a real, well-organized reference bundle. Its one weak spot is conciseness — the body carries lengthy sermons on TDD and prompt-testing philosophy that Claude already knows and that sit oddly in a self-declared index.
Suggestions
Move the generic TDD discipline (red/green/refactor, Given/When/Then, mock-priority ladder) and the 'prompt tests: NEVER assert prose' sermon into a dedicated reference file (e.g. references/testing.md), keeping only the cross-language test-pyramid sizing table and the routing pointer inline.
Trim the seven-rung 'best code is the code never written' ladder and the 'lazy senior engineer' framing to a one-line axiom; the detailed ladder reads as conceptual padding Claude does not need.
Re-read the body against its own 250-LOC ceiling ethos — at ~394 lines it exceeds the limit it enforces; splitting the TDD and code-smells exposition into references would bring the index closer to its stated size discipline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Dense and table-heavy where it counts (ecosystem/toolchain/iron-list tables are token-efficient and project-specific), but the ~394-line body over-explains concepts Claude already knows — the seven-rung 'best code is the code never written' ladder, the full TDD/Given-When-Then/mock-hierarchy sermon, and the long 'prompt tests: NEVER assert prose' section — which is padding for a file that declares itself 'an index'; matches 'mostly efficient but includes some unnecessary explanation'. | 2 / 3 |
Actionability | Highly concrete and copy-paste ready: the awk LOC one-liner, per-language check-script invocations ('uv run scripts/python/check-no-excuse-rules.py', 'cargo +nightly miri test'), exact CI-gate commands, named libraries with versions, and Need→Load routing tables giving exact reference paths; matches the 'fully executable code/commands; copy-paste ready' anchor. | 3 / 3 |
Workflow Clarity | Multi-step processes are clearly sequenced with validation checkpoints: PHASE 0 gate (identify → STOP and read references → apply philosophy), red→green→refactor TDD with 'confirm it fails for the right reason', and a four-step post-write loop whose Step 2 LOC verdict table and Step 3 11-item self-review checklist act as explicit validation with a 'fix before declaring done' feedback loop. | 3 / 3 |
Progressive Disclosure | Genuinely an index: SKILL.md routes via per-language jump tables to references/{python,rust,typescript,go,rust-ub}/README.md plus on-demand files, with code-smells.md and logging.md as cross-cutting refs — all referenced paths verified to exist and clearly signaled one level deep; inline shared-overview content (philosophy, ecosystem tables) is appropriate for the index rather than mis-placed detail. | 3 / 3 |
Total | 11 / 12 Passed |