Content
77%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.
The body is a strong operational contract: commands are executable, the learning loop is sequenced with validation and feedback loops throughout, and edge cases (older binaries, empty stores, TLS/PATH failures) are anticipated. Its weaknesses are size and layout — a ~410-line monolith with repeated guidance that would benefit from splitting the learning protocol and envelope schemas into reference files and trimming duplicated prose.
Suggestions
Move the learning-loop protocol (Steps 1-6 plus the recall envelope JSON) into a references/LEARNING.md and keep a short recall-first summary in SKILL.md, cutting the body roughly in half.
Deduplicate the do-not-interpolate-user-text guidance into one stated-once rule instead of repeating it in Steps 1, 4, and 6, and delete the stray "Public parameter name golden fixture" line at line 32.
Move the path-resolution details and MCP host configuration examples into a references/ reference file, keeping only the doctor and --home essentials inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Most content is CLI-specific contract detail Claude cannot know elsewhere (envelope shapes, exit codes, path resolution), so it largely earns its tokens, but the ~410-line body repeats the do-not-interpolate guidance across Steps 1, 4, and 6, reproduces the full recall envelope JSON inline, and carries a stray duplicate of the description ("Public parameter name golden fixture") at line 32. This matches anchor 3 ("mostly efficient but includes some unnecessary explanation or could be tightened") rather than 4, where trimming would be minor. | 3 / 5 |
Actionability | Every section carries copy-paste-ready commands with concrete flags and example values ("stores find --address example-value ... --agent --select id,name", "npx -y @mvanhorn/printing-press-library install public-param-golden --cli-only", "claude mcp add public-param-golden-pp-mcp -- public-param-golden-pp-mcp"), plus exit-code tables and full JSON envelope examples covering the common cases. Remaining placeholders like <id1> are legitimate parameter slots, not gaps. | 5 / 5 |
Workflow Clarity | The learning loop is a clearly sequenced 6-step protocol with an explicit decision tree, a warnings taxonomy, and feedback loops (learnings confirm prints the payload before materializing; trial-then-confirm; reject tombstones the derivation signature). Validation checkpoints are thorough — verify install and --version before any command, doctor --fail-on warn, --dry-run semantics, and local-data purge previewing before --yes deletion — so the destructive-operation cap does not apply. | 5 / 5 |
Progressive Disclosure | No references/, scripts/, or assets/ directories exist; everything is inlined in a single ~410-line SKILL.md. Section headers are well-organized, but the learning-loop protocol, recall envelope schema, and path/MCP configuration details clearly belong in separate reference files, matching anchor 3 ("content that should be separate is inline"). Not anchor 2 because structure and navigation within the file are good. | 3 / 5 |
Total | 16 / 20 Passed |