Content
87%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The content is concise, highly actionable, and well-structured, but the batch subagent workflow lacks explicit validation and error-recovery checkpoints, which caps workflow clarity at 2.
Suggestions
Add a validation checkpoint after extraction (e.g., confirm each scratch file is non-empty before dispatching agents) and a retry path for failed/empty extractions.
Add a verification step in aggregation (e.g., confirm every subagent returned bullet points and flag empty results) so errors are surfaced rather than silently dropped.
Move the long subagent prompt template into a referenced file if the skill grows, to preserve the lean overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean and executable — bash/jq snippets and a copy-paste prompt template with no padding explaining what CLAUDE.md or conversations are, so every token earns its place. | 3 / 3 |
Actionability | It provides complete executable commands (project-folder discovery, jq extraction, scratch batching) plus a concrete subagent prompt template and explicit batch-size rules, which is copy-paste ready. | 3 / 3 |
Workflow Clarity | The four steps are clearly sequenced, but this is a batch operation over many conversations and there is no validation or error-recovery checkpoint (e.g., verifying agents returned valid findings or handling empty extractions); `2>/dev/null` and `grep -v` silently drop errors, so per the guidelines workflow clarity is capped at 2. | 2 / 3 |
Progressive Disclosure | The skill is under 50 lines, single-purpose, has no bundle files, and is organized into four clearly headed sections, so per the simple-skills note it earns a 3 without external references. | 3 / 3 |
Total | 11 / 12 Passed |