0King0P Skill: Autonomous Self-Correction (S.S.V.E. Loop)
This skill provides a structured method for resolving failures without user intervention, maximizing autonomous reliability.
The S.S.V.E. Loop
1. Stop (Analysis)
Do not touch code immediately. Analyze the error output. Identify the specific file, line number, and error message.
2. Speculate (Hypothesize)
List 2-3 potential causes for the failure.
- Example: "Hypothesis 1: Missing import for
X", "Hypothesis 2: Type mismatch inY", "Hypothesis 3: Environment variableZis unset".
3. Verify (Evidence Gathering)
Use read_file, grep_search, or run_shell_command to gather evidence for your first hypothesis.
- Validation: Do NOT apply a fix until you have seen the evidence (e.g., you've read the file and confirmed the import is indeed missing).
4. Execute (Targeted Fix)
Apply the fix only for the verified hypothesis. Use the robust-edit protocol.
5. Evaluate (Re-verification)
Re-run the failing command.
- Success: Update the handoff and proceed.
- Failure: Repeat the loop with a new hypothesis. If 3 loops fail, stop and escalate to the user with your findings.
Best Practices
- No Blind Retries: Never apply the same fix twice hoping for a different result.
- Isolate Changes: Fix one error at a time.
- Root Cause: Aim to fix the source of the error, not just the symptom.