Verifiedon 2026.7.1-2
Action boundary
Before you act
- Expected result
- A timestamped diagnostic packet proves classification, bounded checks, one repair or escalation, validation, rollback condition, and accountable handoff.
- Failure mode
- The learner makes several speculative changes, pastes secrets/log transcripts into evidence, or declares success without a post-check.
- Rollback
- Restore the one documented change where applicable; otherwise leave the sandbox unchanged and escalate with the original evidence packet.
Scenario
Use an instructor-provided sandbox fault: invalid configuration, unavailable gateway, degraded channel account, unavailable model, or retained dead letter. Do not plant faults in production. Start a ten-minute timer and state the symptom before typing a command.
Rules of engagement
- Collect version and help evidence.
- Use non-destructive checks first: config validation, doctor, health/status, scoped channel probe, and bounded logs as relevant.
- Classify the failure layer and name one hypothesis.
- Make at most one approved, reversible change. A no-change escalation is valid.
- Run the post-check that would disprove your original diagnosis.
- Keep evidence redacted; no secrets, payloads, recipient identifiers, or unbounded logs.
Deliverable
Submit the handoff template from the previous lesson plus a visual fault map. The map must show symptom → evidence → classification → change/no-change → post-check → rollback or owner. Label uncertainty rather than filling gaps with assumptions.
Rubric
| Criterion | Weight | Pass signal |
|---|---|---|
| Evidence quality | 35% | Version, timestamps, bounded command results, and source of truth are present. |
| Least-change repair | 25% | One change is justified by the observed failure, or escalation preserves state. |
| Rollback condition | 20% | A concrete stop/revert condition and owner are named before mutation. |
| Clear handoff | 20% | Another operator can reproduce the checks without receiving secrets or private content. |
A pass is not a claim of production readiness. It is proof that you can reduce ambiguity without widening blast radius.
Check your understanding
Source provenanceVerification and sources
Review receipt rr_cli_diagnostic_capstone
- Outcome
- approved
- Method
- source-review
- Reviewer
- forge-independent-review
- Reviewed
Evidence
- openclaw-docs-index — openclaw-main-1237c0c-cli-health; snapshot
51cc0e3dd61a…
Limitations
- Primary-source editorial review; no production diagnostic was performed.
Lesson checkpoint
Ready to move on?
Mark this lesson complete when you can apply its outcome without relying on the examples above.