grtninja/request-loopback-resume

Request Loopback Resume

Use this skill when you need to return to a prior request without losing execution intent.

Workflow

1. Initialize a workstream state file with task + lanes.
2. Update lane status, next-step notes, and evidence artifacts during work.
3. Validate state before resuming (single in-progress lane rule).
4. Generate deterministic resume action (continue, start, blocked, done).

Canonical Commands

Initialize:

python3 "$CODEX_HOME/skills/request-loopback-resume/scripts/workstream_resume.py" \
  init \
  --state-file .codex/workstreams/example.json \
  --task "Upgrade media pipeline skills" \
  --lane "skill update" \
  --lane "validation" \
  --lane "docs"

Update checkpoint:

python3 "$CODEX_HOME/skills/request-loopback-resume/scripts/workstream_resume.py" \
  set \
  --state-file .codex/workstreams/example.json \
  --lane-status "skill update=completed" \
  --lane-status "validation=in_progress" \
  --lane-next "validation=Run arbiter + installer evidence" \
  --artifact /tmp/repo-b-media-skills-arbiter.json

Resume action:

python3 "$CODEX_HOME/skills/request-loopback-resume/scripts/workstream_resume.py" \
  resume \
  --state-file .codex/workstreams/example.json \
  --json-out /tmp/workstream-resume.json

Status Model

Allowed lane status values:

• pending
• in_progress
• completed
• blocked

Rules:

1. At most one lane can be in_progress.
2. resume action is continue for current in_progress lane.
3. If no lane is in_progress, action is start for first pending lane.
4. If only blocked/completed lanes remain, action is blocked or done.

Checkpoint Discipline

Use deterministic checkpoint cadence:

1. Write checkpoint after each lane transition (pending -> in_progress -> completed/blocked).
2. Attach artifact paths for each completed lane.
3. Keep one clear lane-next value for the current lane.
4. Re-run resume after any interruption before executing further commands.

Resume Evidence Contract

Preserve:

• state_file
• latest resume_json
• current lane_status map
• latest artifact paths

If any element is missing, resume should fail closed and reroute through skill-hub.

Scope Boundary

Use this skill for resuming prior user requests and preserving deterministic continuity across interruptions.

Do not use this skill for:

1. Repo-specific implementation/debugging lanes.
2. Skill admission/classification decisions.
3. Release version/changelog operations.

References

• references/state-contract.md
• scripts/workstream_resume.py

Loopback

If this lane is unresolved, blocked, or ambiguous:

1. Capture latest resume JSON output.
2. Route through $skill-hub for chain recalculation.
3. Continue only after a deterministic next lane is selected.