Content
81%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-engineered dual-mode skill body: complete sequenced workflows with decision gates and validation checkpoints, concrete commands and a copy-ready threshold table, and clean one-level reference splitting. The gaps are executional rather than structural — placeholder data payloads, a referenced ledger.py script missing from the bundle, and cross-bundle paths that don't resolve from this directory.
Suggestions
Bundle scripts/connectors/ledger.py (or inline the minimal record/diff/trend usage) and replace the '"sessions": ...' ellipsis payload with one complete example JSON so the zero-dependency measurement loop is copy-paste executable.
Make the out-of-bundle paths resolvable: either vendor the referenced sections of ../../../references/measurement-protocol.md and skill-contract.md into the skill's own references/, or summarize their required content inline so the readback protocol is self-contained.
State the Measured / User-provided / Estimated labeling rule once (e.g., in the Shared discipline section) and reference it from the alert and steep-decline sections instead of repeating the full rule four times.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body never explains concepts Claude already knows (no 'what SEO is' padding) and the threshold table packs concrete values densely, but the Measured/User-provided/Estimated labeling discipline is repeated four times (Data Sources, Instructions, alert step 2, steep-decline trigger) and protocol phrasing like 'bind the current non-forked page/content/change head' could be tightened. Efficient with minor trimmable repetition — the 4 anchor, not 5. | 4 / 5 |
Actionability | Concrete executable commands are present ("python3 "${CLAUDE_PLUGIN_ROOT}/scripts/connectors/ledger.py" record <domain> --source monitor --data '...'", "ledger.py diff", "ledger.py trend") plus exact thresholds (-15% WoW, >3 position drop) and Quick Start invocations. Falls short of the 5 anchor because the data payload uses ellipsis placeholders ("{"sessions": ..., ...}") and the ledger.py script is not actually present in the bundle's scripts/ directory, leaving minor execution gaps. | 4 / 5 |
Workflow Clarity | Step 0 mode selection feeds two fully numbered procedures (report steps 1-11, alert steps 1-5) with explicit stop-gates ("Mode ambiguous", missing period, missing baseline) and continue-silently rules, done-when checklists, and validation discipline ("confirm the drop holds before declaring an incident", "corroborate before stating a cause"). Clear sequence with explicit checkpoints and error-recovery options — matches the 5 anchor; this is not a destructive/batch skill so no cap applies. | 5 / 5 |
Progressive Disclosure | Good structure: all six references/*.md files exist, are one level deep, cited at point of use, and indexed with one-line descriptions in a Reference Materials section. Falls short of 5 because several load-bearing paths point outside the bundle and do not resolve — ../../../references/skill-contract.md, ../../../references/measurement-protocol.md (which the readback section depends on), ../../../CONNECTORS.md, and the absent scripts/connectors/ — creating navigation gaps. | 4 / 5 |
Total | 17 / 20 Passed |