Content
82%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-structured, actionable skill with concrete SQL and tool examples and a clear multi-step workflow. Slight room to tighten prose and offload reference material for better token efficiency.
Suggestions
Trim framing sentences like 'Start with the recording to get metadata and the person's distinct ID' to lean imperatives to lift conciseness toward 5.
Consider moving the extended Replay Vision scanner-creation procedure into a referenced file (e.g. VISION.md) so the core investigation workflow stays scannable.
Add an explicit verification step after the parallel queries in Steps 1–3 (e.g. confirm distinct_id resolved and events returned) to add a light checkpoint to the main workflow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Efficient and assumes Claude's competence — no padding about what PostHog or a session recording is — but a few framing sentences ('Start with the recording to get metadata...') could be trimmed. | 4 / 5 |
Actionability | Provides fully executable, copy-paste-ready tool calls and SQL queries with parameter placeholders covering the common investigation cases. | 5 / 5 |
Workflow Clarity | Clear Step 1–4 sequence with conditional logic and a polling feedback loop in the Vision section, but the core read-only steps lack explicit validation checkpoints. | 4 / 5 |
Progressive Disclosure | Well-organized with clear sections (tools, workflow, tips, related skills) and no nested references, though the ~230-line body is somewhat long for a single file with no offloaded reference material. | 4 / 5 |
Total | 17 / 20 Passed |