Rocky Incident Notes: Capture, Diagnose, Prevent

Incidents happen: a deploy fails, a route renders poorly, a provider rejects a request, or an automation produces the wrong artifact. A good incident note turns the failure into a safer next run without exposing private data.

Capture the facts

• Timestamp and affected public route or artifact.
• User-visible symptom.
• Command or check that failed.
• Whether production was affected.
• Screenshots or logs only after removing secrets and private identifiers.

Diagnose before fixing

• Reproduce the smallest failing case.
• Separate local failure, build failure, deploy failure, and live rendering failure.
• Check recent changes before broad refactors.
• If a public page looks wrong, use screenshot/contact-sheet QA rather than only HTTP checks.

Turn the fix into prevention

• Add a test or script check when possible.
• Add a checklist item for visual issues that scripts cannot catch.
• Patch a reusable Hermes skill when the lesson affects future work.
• Avoid storing stale one-off details as long-term memory.

Pitfalls

• Writing blame instead of evidence.
• Including tokens, local usernames, customer details, or private project paths.
• Declaring recovery after a build passes but before the live URL is checked.
• Forgetting to remove temporary QA artifacts from deployment context.

Verification steps

• Re-run the failing check.
• Run a broader smoke test.
• Verify the live route or artifact.
• Summarize root cause, fix, and prevention in a client-safe way.