Repair

qp repair reruns guard checks and produces structured remediation context designed for fast fixes and agent handoff.

What Repair Does

For a guard run, repair collects:

1. overall guard status
2. failing steps
3. parsed diagnostics (when available)
4. scope metadata for failing tasks
5. current git diff snippet (full mode)
6. suggested next action text

Basic Usage

qp repair
qp repair ci

If no guard name is provided, repair targets the default guard flow.

Output Modes

Full markdown report

qp repair ci

Includes guard status table, per-failure details, truncated git diff, and next-action guidance.

Brief handoff mode

qp repair ci --brief

Produces concise failure-centric output for quick copy/paste into issue comments or agent prompts.

JSON mode

qp repair ci --json

Returns structured fields:

• guard, overall, exit_code, ran_at
• failures[] (task, status, scope, parsed errors, stderr)
• git_diff
• suggested_next_action
• rendered markdown

Clipboard mode

qp repair ci --copy

Copies markdown output directly to clipboard.

Unsafe Task Handling

If repair guard steps include destructive/external tasks:

qp repair ci --allow-unsafe

This mirrors guard safety behavior and keeps risky execution explicit.

Worked Example

tasks:
  lint:
    desc: Lint code
    cmd: golangci-lint run
    scope: backend

  test:
    desc: Run tests
    cmd: go test ./...
    error_format: go_test
    scope: backend

guards:
  ci:
    steps: [lint, test]

qp repair ci --brief

Brief output highlights failing task names, scope paths, and parsed error entries so fixes can start immediately.

Repair + Planning Loop

A practical loop:

1. qp repair ci --json
2. extract failing files/scopes
3. qp agent-brief --file ... or qp diff-plan
4. apply fix
5. rerun qp repair ci

This keeps remediation grounded in observed failures and bounded change areas.

Next Step

For end-to-end collaboration patterns with coding agents, continue to Agent Integration.