Content
86%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 high-quality, deeply actionable scout skill with executable SQL, precise thresholds, and a clear sequenced workflow; its only soft spots are minor framing verbosity and the absence of explicit error-recovery retry loops (less critical for a read-only reporting skill).
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Long but mostly efficient — it assumes Claude knows SQL and PostHog and adds only replay-specific footguns and judgment, with no padding about what a session replay is; a few framing sentences ('Two mechanical facts anchor everything') could be trimmed, keeping it just below the lean-and-efficient anchor. | 4 / 5 |
Actionability | Every step ships executable SQL with exact table and column names (raw_session_replay_events.min_first_timestamp, argMinMerge(first_url)), precise volume gates (~3× prior-13-day mean, ≥10 sessions, ≥5 persons), and exact tool names with parameters — fully copy-paste ready. | 5 / 5 |
Workflow Clarity | A clearly sequenced run (orient → explore → decide → close out) with embedded validation gates (baseline requirements, volume gates, 'require corroboration before emitting'), but no explicit validate→fix→retry feedback loops; this sits at clear-sequence-with-minor-gaps rather than the explicit-checkpoint-recovery anchor. | 4 / 5 |
Progressive Disclosure | Well-organized with clear section headers and a single one-level-deep reference ('authoring-scouts → references/report-contract.md') that is explicitly signaled and explicitly excluded from restatement; the inlined SQL is the core skill content and belongs in the body, so navigation is easy. | 5 / 5 |
Total | 18 / 20 Passed |