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 a high-signal, actionable debugging guide with clear sequenced workflows, validation checkpoints, and well-structured one-level-deep references to real bundle files. Its only weakness is mild verbosity from restated table content and slightly long prose sections, which keeps conciseness below the top anchor.
Suggestions
Collapse the 'Consequences worth memorizing' bullets into references to the parity table above, or trim them to only the non-obvious consequences not already readable from the table, to remove restated content.
Tighten the Repos and Access sections by leading with the commands/endpoints and moving rationale into shorter clauses, cutting prose that assumes less than Claude's competence.
Consider promoting the detailed 'How a survey actually gets shown' pipeline walkthrough to a reference file if it grows, keeping the inline version to the ordered checklist plus the two non-obvious facts.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and packed with PostHog-specific knowledge Claude lacks, but the 'Consequences worth memorizing' bullets restate the parity table and the repos/access prose runs long enough to tighten. Not padded with basic concepts, yet not lean enough for the top anchor. | 2 / 3 |
Actionability | Provides executable commands (scripts/repos.py invocations), concrete API endpoints with parameter shapes, exact file paths to grep, and links to copy-paste-ready HogQL templates — fully actionable guidance. | 3 / 3 |
Workflow Clarity | The 7-step debugging workflow has an explicit disambiguation checkpoint ('none vs fewer') and a 'confirm with one targeted query' gate, while each known cause pairs a 'Signature' with a 'Fix', giving clear sequencing with validation feedback loops. | 3 / 3 |
Progressive Disclosure | Keeps core tactical overview inline while externalizing separable material into verified one-level-deep references (diagnostic-queries.md, local-repos.md) and a script (repos.py), each clearly signaled with a description of its contents. | 3 / 3 |
Total | 11 / 12 Passed |