Content
67%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, command-first operational guide: exact invocations for every subcommand, explicit decision rules for vertical vs. general search, ordered platform detection, and genuine error-recovery loops. Its main weaknesses are localized redundancy (duplicated runtime tables and key-persistence steps) and two references to files (SHA256SUMS.txt, requirements.txt) that are absent from the bundle.
Suggestions
Collapse the duplicate runtime tables into a single CLI invocation table (the doc command row can say 'same invocation with the doc subcommand') and merge the 'Persisting Keys' section into the Scenarios table row to remove the repeated four-step procedure.
Either add SHA256SUMS.txt and requirements.txt to the bundle or delete the instructions referencing them, so every written step is executable.
Add brief guidance for handling per-item failures in batch_search responses (detect, retry individually, or fall back to sequential search) to complete the batch-operation feedback loop.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and operational (cheat sheet, rules, tables) with no educational padding, but contains clear duplication: the runtime invocation table appears twice with near-identical rows (doc table and CLI Invocation table), and the key-persistence procedure is stated twice (Scenarios table row and the 'Persisting Keys' section repeat the same four steps). Fits 'mostly efficient but could be tightened'; not 2 because there is no explanation of concepts Claude already knows and the padding is localized redundancy rather than pervasive verbosity. | 3 / 5 |
Actionability | The Command Cheat Sheet gives exact, copy-paste-ready command shapes for search, batch_search, extract, and get_sub_domains, including both --sdp formats and explicit invalid examples, which approaches fully executable. However, two written instructions cannot be executed against the actual bundle: 'verify the bundled CLI scripts locally against SHA256SUMS.txt' and 'pip install -r requirements.txt' reference files that do not exist in the bundle — minor gaps that fit the 4 anchor rather than 5. | 4 / 5 |
Workflow Clarity | Sequences are explicit and well-ordered: runtime.conf fast path, a numbered platform-detection procedure with first-success logic, a MUST rule to call get_sub_domains before vertical search, and error-recovery feedback loops (runtime fallthrough, retry after key save, --help on argument errors). Not 5 because batch_search results have no per-query failure validation guidance (e.g., how to detect and retry a failed item in a batch response); not 3 because validation and feedback loops for CLI/key failures are explicitly present. | 4 / 5 |
Progressive Disclosure | Scored against the actual bundle: the full interface reference is correctly kept out of SKILL.md behind the offline doc command, which genuinely reads scripts/shared/doc_spec.md (verified in anysearch_cli.py), and referenced CLI script paths all exist — a good one-level-deep structure. Not 5 because navigation relies on the indirect doc command rather than clearly signaled reference files, and SHA256SUMS.txt/requirements.txt are mentioned but missing from the bundle. | 4 / 5 |
Total | 15 / 20 Passed |