Conciseness

The body explains nothing Claude already knows; every paragraph carries non-obvious internal specifics ("Design's table alone is over 1 GB, and an unfiltered scan times out against production", "Their conversation lives in that client; only the calls reached us"). Lean throughout with no trimmable over-explanation.

Actionability

Two copy-paste commands with flags ("pnpm exec tsx scripts/debug-runs.ts find <input> [--app NAME] [--days 30] [--limit 20]"), an input-matching table with concrete examples for every input type, and ordered diagnostic steps ("Start at the failed run's `error`, then the last tool call before it"). Fully executable guidance covering the common cases.

Workflow Clarity

Clear numbered sequence (Find the runs → Read a thread → Diagnose) with explicit error-recovery loops: "`auditError`: the audit scan timed out; rerun with a smaller `--days`", the error-entry disambiguation ("it was not searched, which is not the same as 'nothing found'"), and the fallback "Threads created before May 2026 have no run ids, so `runs[]` is empty and only the transcript is available". The workflow is read-only, so the destructive/batch validation cap does not apply.

Progressive Disclosure

A single well-organized file with clear sections (Find / Read a thread / Diagnose / Retention / Rules) and nothing inlined that belongs in a separate file, but its sole external reference — `scripts/debug-runs.ts`, invoked by every command — is not present in the bundle (no scripts/ directory exists), so the reference does not resolve. Not 5 because a clearly signaled one-level-deep reference is broken relative to the provided bundle; not 3 because structure and signaling are otherwise good.

Specificity

"Find and read the agent chat runs behind a run id, a user's email, or an object id or link … then work out what failed or why the object looks the way it does" names several concrete actions with enumerated input types and object kinds. Not 3 because more than 1–2 concrete actions are given with only minor gaps; not 5 because the action list is not comprehensive (e.g., reading full transcripts or listing versions/external sessions are unnamed capabilities).

Completeness

Both halves are explicit: the what is "Find and read the agent chat runs … then work out what failed or why the object looks the way it does", and the when is "Use when someone reports a broken chat, a failed run, or a bad result and gives you any of those instead of a thread". This matches the anchor with concrete trigger phrases on both sides; not 4 because the when-clause is already explicit and specific rather than merely present.

Trigger Term Quality

Natural reporter phrases are present — "broken chat", "failed run", "bad result" — alongside concrete trigger forms (run id, email, object id or link; deck, design, dashboard, document, plan, recording). Good coverage, but common synonyms such as "stuck", "wrong output", or "debug this run" are missing, so it falls short of the comprehensive-synonym anchor at 5.

Distinctiveness Conflict Risk

"agent chat runs behind a run id, a user's email, or an object id or link … in any app" carves out a clear niche, and the trigger is tightly bounded by "gives you any of those instead of a thread", so it is unlikely to fire for unrelated debugging or document skills. Minimal overlap risk.