Content
96%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 dense, highly actionable guide that assumes Django knowledge and spends every token on PostHog-specific facts: exact file paths, class names, throttle cache keys, a four-step wiring checklist, and a concrete test matrix. The only gap is structural — no reference files at all, with a ~95-line body that slightly exceeds ideal single-file size.
Suggestions
Move the minimum test-matrix ('Cover at minimum: ...') into a references/testing.md file and reference it one level deep, keeping the checklist in SKILL.md lean.
Consider relocating the 'What you get for free' details (query tagging internals, last_used_at mechanics) to a short reference file, since they are context rather than wiring steps.
Trim the curl example's '-d '{...}'' placeholder to a concrete JSON body for the planned 'run' action consumer to make it fully copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Every section carries project-specific facts Claude cannot know (file paths, cache-key formats like 'psak:{key_id}', default-deny behavior, test file locations) and assumes Django competence; there is no explanation of general concepts and no padding. Not level 4: there are no sections that could be trimmed without losing information. | 5 / 5 |
Actionability | Concrete, executable guidance throughout: the scope allowlist tuple, the viewset wiring snippet ('authentication_classes = [ProjectSecretAPIKeyAuthentication]', 'psak_allowed_actions = ["run"]'), named throttle classes, permission helpers, and a curl example. Placeholders like '<your_action_path>' are inherent variables, not pseudocode, so the top anchor fits. | 5 / 5 |
Workflow Clarity | A numbered four-step 'all required' checklist with per-step failure modes ('key creation rejects the scope before auth is ever attempted') plus an explicit testing section enumerating expected outcomes (200 vs 403 cases), serving as the validation checkpoint. Not level 4: validation expectations are explicit and complete for each failure mode. | 5 / 5 |
Progressive Disclosure | Sections are well-organized and navigation is easy, but the ~95-line body is above the under-50-line single-file exception and ships no bundle files; content like the test matrix and the 'What you get for free' details could live in one-level-deep references. Not level 5 because there are no well-signaled reference files to split content into; not level 3 because structure and organization are genuinely good. | 4 / 5 |
Total | 19 / 20 Passed |