Conciseness

The body is lean and fully operational — exact log signatures, an env-toggle table, and copy-paste commands with no padding or explanation of concepts Claude already knows; the one justificatory note ("Why not pyspy/gdb/strace? Apps pods are locked down") preempts a wrong alternative rather than over-explaining.

Actionability

Guidance is fully executable: `https://<app-url>/logz`, `databricks apps logs <app-name> --follow --profile <profile>`, the `jq '.app_status'` status check, an `app.yaml` env snippet, and `kill -USR1 <pid>`, covering SSO/non-SSO, Python/Node, and crash/hang cases.

Workflow Clarity

The sections sequence the investigation clearly (get logs, with explicit SSO vs CLI branching → match the symptom-to-cause signatures in "What to look for" → hang procedure with explicit prioritization "check that first"). Diagnosis is read-only, so the destructive/batch validation cap does not apply, and the single investigation path is unambiguous.

Progressive Disclosure

No bundle files exist (verified: no references/, scripts/, or assets/ directories) and none are needed at this scope; the single file has well-organized sections and its one external pointer (".scripts/agent-integration-tests/crash-examples/") is clearly signaled and one level deep.

Specificity

The description names the domain and several concrete capabilities ("Diagnose a crashed, hung, or erroring deployed Databricks App", stack traces, reading app logs, diagnostics env toggles), which sits at the 'several specific actions; minor gaps' anchor; however, the rubric's second-person penalty applies because trigger (3) uses "you need the stack trace behind a failure", reducing the score from 4 to 3.

Completeness

It explicitly answers both questions: "Diagnose a crashed, hung, or erroring deployed Databricks App" states the what, and "Use when: (1)…(4)" gives four concrete trigger conditions for the when. It clearly matches the anchor requiring explicit what AND when with concrete trigger phrases.

Trigger Term Quality

The triggers are the literal symptom phrases a user would echo ("crashed / UNAVAILABLE / 502s", "hangs or is unresponsive", "stack trace", "app logs") with synonym clusters covering both crash-like and hang-like states; no meaningful natural terms are missing for this domain.

Distinctiveness Conflict Risk

"deployed Databricks App" plus platform-specific triggers (UNAVAILABLE/502s, diagnostics env toggles) carve out a clear niche with minimal overlap risk against other skills; nothing generic remains that would fire the wrong skill.