Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is highly actionable with executable tool calls, scripts, and SQL, a clear sequenced debug workflow, and well-signaled one-level-deep references. The main weakness is conciseness: the parsing scripts and key warnings are restated multiple times and could be consolidated.
Suggestions
Consolidate the parsing-script coverage: keep the inline bash block in Step 3 OR the full table under "Parsing large trace results", and cross-reference the other rather than maintaining both.
State the posthog.ai_events heavy-content rule once in the body (e.g. in the Tips or event reference link) and link out for detail instead of repeating it across the description, Step 2, and the Tips section.
Reduce repeated _posthogUrl reminders to a single authoritative callout (e.g. one Tip bullet) and reference it where needed.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and free of concept-explanation fluff, but the parsing scripts are duplicated (inline bash block in Step 3 and again as a full table under "Parsing large trace results"), and the posthog.ai_events and _posthogUrl points are repeated several times, so it could be tightened. | 2 / 3 |
Actionability | Provides fully executable JSON tool-call payloads, concrete bash invocations with env vars, and complete HogQL/SQL examples that are copy-paste ready, matching the anchor for executable, specific guidance. | 3 / 3 |
Workflow Clarity | The URL-to-debug workflow is a clearly sequenced three-step process, and the schema-discovery-first pattern acts as an explicit pre-filter checkpoint; operations are read-only so the destructive-operation feedback-loop cap does not apply. | 3 / 3 |
Progressive Disclosure | SKILL.md is a clear overview pointing one level deep to real bundle files (events-and-properties.md, example-llm-trace.md.j2, example-llm-traces-list.md, scripts/*) with a navigable scripts table; reference cross-links are sibling-level rather than deeply nested. | 3 / 3 |
Total | 11 / 12 Passed |