| name | foreman-safe-recovery |
| description | Use when deciding between Foreman retry, reset, doctor --fix, worktree clean, abandon, or clean-state for failed/stuck/obsolete runs, branches, worktrees, or PRs. |
Foreman Safe Recovery
When to Use
Use this skill for failed/stuck tasks, obsolete work, stale worktrees, retryable failures, branch cleanup, stale runtime assets, or operator cleanup.
Recovery Decision Tree
- Diagnose first with
foreman-pipeline-diagnosis order.
- Use dry-run options before destructive commands when available.
- Use
foreman retry <task-id> --dispatch for understood retryable failures.
- Use
foreman reset <task-id> --dry-run before reset; add --keep-worktree when preserving uncommitted investigation matters.
- Use
foreman doctor --fix for safe automated stale prompt/workflow/skill/worktree checks/fixes.
- Use
foreman worktree clean instead of manual rm -rf for stale worktrees.
- Use
foreman abandon <task-id> for obsolete work that should not land.
- Use
foreman clean-state --dry-run before --force for state cleanup.
Evidence to Record
Record the task ID, run ID, failed phase, current status, root cause classification, command chosen, dry-run result, and reason string.
Guardrails
- Do not reset/abandon merged work as routine cleanup.
- Do not delete branches/worktrees when important work exists only in working copy.
- Do not use reset to hide a real implementation or verification failure.
