Ships in the template with a safe conservative default. Change via PR for a full audit trail.
# Change via PR — the commit history is your governance record
# vars.COCO_MAX_AUTO overrides this value for runtime experiments
fix_mode:
max_auto: conservative # aggressive | conservative | off
# aggressive: auto-fix when AI confidence is medium or high
# conservative: auto-fix only when severity=low, complexity=low, confidence=high (default)
# off: never auto-fix — always require /coco fix commentSet in GitHub Actions repository variables (Settings → Secrets and variables → Actions → Variables) or as a GitLab CI/CD variable (non-masked).
The variable takes priority over the config file. Use it to run experiments without committing a policy change:
# GitHub — set via CLI
gh variable set COCO_MAX_AUTO --repo "$REPO_PATH" --body "aggressive"
# GitLab — set via API
glab api "projects/$ENCODED_PATH/variables" --method PUT \
-F "key=COCO_MAX_AUTO" -F "value=aggressive"Remove the variable to revert to the config file value.
1. vars.COCO_MAX_AUTO ← runtime experiment
2. .github/coco-config.yml ← config-as-code default
3. Built-in: "conservative"
Every scan run logs the active ceiling and source to the Actions/pipeline summary:
::notice::Fix ceiling: aggressive (source: vars.COCO_MAX_AUTO override)
::notice::Fix ceiling: conservative (source: .github/coco-config.yml)
!!! tip "Governance tip"
Keep max_auto: conservative in the config file permanently.
Use vars.COCO_MAX_AUTO only for time-boxed experiments.
When the experiment ends, delete the variable — policy reverts to the reviewed default.