Content
75%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.
An impressively lean, instruction-only body with a sensible implicit workflow, explicit safety gates (sanitize evidence, never repeat destructive actions, draft-only unless authorized), and honest-reporting checkpoints. Its main weakness is actionability: how to actually query screenpipe-api and what shape the final report should take are left unspecified.
Suggestions
Add one concrete example of inspecting recorded activity with screenpipe-api (e.g., a specific query for frames around the failure time) so the investigation step is executable rather than a hint.
Include a minimal bug-report skeleton (title, environment, steps to reproduce, expected/actual, evidence links) so the output format is unambiguous.
Add short section headers for the three phases (Investigate / Write / Sanitize & deliver) to make the implicit sequence and safety gates scannable.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Roughly 150 words, three lean paragraphs, zero padding or explanation of concepts Claude already knows — every sentence instructs ('Resolve the affected app...', 'Remove credentials...', 'Do not repeat destructive actions'). Matches 'Lean and efficient; assumes Claude's competence; every token earns its place'; not a 4 because there is no over-explanation to trim. | 5 / 5 |
Actionability | Concrete deliverables are enumerated (reproduction steps, expected behavior, actual behavior, environment/version, evidence links) plus specific sanitization items ('credentials, unrelated screen content, personal identifiers, and private business metrics'), but key execution details are missing: no example of how to inspect activity/frames with screenpipe-api and no report format or template. Fits 'Some concrete guidance but incomplete... missing key details'; not a 4 because 'inspect the relevant activity and frames' is a high-level hint rather than executable guidance. | 3 / 5 |
Workflow Clarity | A clear three-phase sequence (resolve context and recover the sequence via screenpipe-api; write the report; sanitize and return a draft) with several explicit checkpoints: 'environment/version when verified', 'Mark steps that could not be reproduced', 'Return a draft report unless issue creation is explicitly authorized'. Not a 5: the phases are implied by paragraph order rather than enumerated, and there is no review/fix loop on the draft; not a 3: validation checkpoints are explicitly present, and the workflow is read-only rather than destructive or batch, so no cap applies. | 4 / 5 |
Progressive Disclosure | Under 50 lines with no bundle files and no content that belongs in separate references, so the structure is appropriate for a small skill; the three paragraphs are thematically ordered (investigate / write / sanitize). Not a 5: the body has only a single '# bug report' heading with no section headers, so it is not organized into well-signaled sections; not a 3: nothing is inlined that should be split out, and navigation over content this short is unambiguous. | 4 / 5 |
Total | 16 / 20 Passed |