flame-harness-resume

flame-harness-resume

Resume handler for the flutter-flame-harness pipeline. Reads state.md, branches on pause_reason, and re-dispatches the next phase skill when ready to continue.

All file schemas (state.md, pipeline-log.md) and state transition rules (including the requirement that pause_reason is cleared to "" when status is restored to running) are defined in docs/harness-protocol.md — refer to §2 (state.md schema) and §7 (state transition rules) as the single source of truth. Do not redefine schemas here.

---

Input

Read docs/harness/state.md per the schema in docs/harness-protocol.md §2. If state.md does not exist, abort with:

flame-harness-resume: nothing to resume — docs/harness/state.md not found.
Run /flame-harness <idea> to start a new pipeline.

Extract:

• status — expected paused; if running or completed, print a note and exit (nothing to resume).
• pause_reason — rate_limit | manual_action | error
• next_role — role to dispatch after resuming
• resume_attempts — integer; will be incremented on successful resume
• current_phase, updated_at — for logging

If status is running:

flame-harness-resume: pipeline is already running (status: running, next_role: <next_role>).
No resume needed.

If status is completed:

flame-harness-resume: pipeline is already completed. Nothing to resume.

---

Branch on pause_reason

rate_limit

pause_reason: rate_limit means the pipeline hit an API rate limit and paused automatically.

Procedure:

1. Print:

   Pipeline paused due to rate limit (rate_limit).
   Checking whether the rate-limit window has passed...
   

2. Read updated_at from state.md. Calculate elapsed time since that timestamp using Bash (date command). If elapsed time is less than 60 seconds, print a warning:

   Warning: only <N> seconds have elapsed since the pause. The rate-limit window may not
   have passed yet. Proceeding anyway — if another rate limit is hit, the pipeline will
   pause again.
   

3. Proceed to Resume execution below.