Skip to content

T-JWCO-orchestrate-first-live-tick-shakedown

Status: closed/done · Impact: high · Complexity: small

Auto-generated from a /sdlc:task-work post-mortem. Review and promote to open/ready before picking up.

T-3OVF-add-orchestrate-skill shipped as spec-only — AC-1 / AC-3 / AC-6 were all verified by dry-run trace, not by an actual tick. The hard parts (Agent dispatch shape, pr-check / task-work return parsing, digest-log writing, parallelism cap accounting) are unexercised. The expected outcome of a first real run is at least one prose-vs-implementation gap that only shows up under live conditions. Closing this gap means deliberately running /loop /sdlc:orchestrate against the live repo for a small number of ticks, recording what broke, and either fixing the prose in orchestrate’s SKILL.md or filing follow-ups.

From the originating post-mortem:

The skill has not been exercised live yet — AC-1 / AC-3 / AC-6 are all dry-run traces, not actual ticks. A first /reload-plugins

  • /loop /sdlc:orchestrate run would surface any prose-vs- implementation gaps. The hard parts (Agent dispatch shape, pr- check return parsing, task-work return parsing) are spec-only until then.

There is no entry in the project for “run orchestrate live and capture findings”. Without one, the spec-only ACs in the originating task will silently stay spec-only.

A deliberate shake-down run: invoke /reload-plugins then /loop /sdlc:orchestrate against the current repo state, let it run for at least 3 ticks (or however many are needed for a non-trivial state transition), capture the digest log, and write a short post- run note. Any findings either become inline fixes to plugin/skills/orchestrate/SKILL.md (small prose patches) or new follow-up tasks.

  1. Snapshot the repo state pre-run (open PRs, ready tasks, in- flight worktrees) so the digest can be cross-checked.
  2. /reload-plugins, then /loop /sdlc:orchestrate. Let it run 3+ ticks.
  3. Read .claude/orchestrator-log.md and verify each tick’s declared actions actually happened (the matching sub-agents were dispatched, PRs were triaged, tasks were picked up).
  4. Note any discrepancies; for each one decide: patch SKILL.md in place (small wording fix) or file a follow-up task.
  5. Flip the spec-only ACs (AC-1, AC-3, AC-4, AC-6) in T-3OVF-add-orchestrate-skill from agent-manual claims to verified — or, if any cannot be verified yet, surface why.
  • .claude/orchestrator-log.md — generated artifact, not checked in. The shake-down reads it.
  • plugin/skills/orchestrate/SKILL.md — minor prose fixes only, if needed.
  • docs/planning/tasks/2026-05-19-add-orchestrate-skill.md — AC state updates post-run (separate commit; the originating task is already merged by then).
  • AC-1: A digest-log file at .claude/orchestrator-log.md exists with at least 3 tick entries from the live run.
  • AC-2: Each tick’s recorded actions are cross-checked against ground truth (PR list, task statuses, worktrees) and any divergences are filed as follow-up tasks or fixed in SKILL.md.
  • AC-3: At least one of the orchestrator’s spec-only ACs is promoted to “verified” with evidence cited from the digest log.

Spawned by /sdlc:task-work post-mortem of T-3OVF-add-orchestrate-skill on 2026-05-20.


← Back to Tasks