full-pipelinelisted

# Bridge: Full Pipeline Orchestrator **Announce at start:** "I'm using the bridge:full-pipeline skill to run the workflow end-to-end, chaining each step's output into the next." ## Purpose This skill doesn't do any of the work itself — it's a sequencer. Each numbered step below is a real skill invocation; this skill's only job is to run them in order, carry the right file/output forward as the next step's input, and stop to surface a decision back to the user whenever a step can't proceed without one (e.g. `autoresearch-plan` finding no real alternatives to compare, or a review step failing). Don't skip a step to save time, and don't silently substitute a different skill for the one named — if the user wants a different implementer or reviewer mid-run, that's a decision to surface, not one to make unilaterally. Because every step runs in the same conversation, this pipeline accumulates context as it goes. For a long-running or multi-day pipeline, it's fine — expected, even — for the conversation to pause between steps and resume later; nothing here requires unattended execution. ## Sequence 1. **`/office-hours`** (gstack) — clarify the problem, constraints, and rough direction before any plan exists. 2. **`/autoplan`** (gstack) — CEO/design/eng triple review, produces a locked, reviewed plan. 3. **Database checkpoint A** — see "Database checkpoint" below, run against the plan from step 2. 4. **`bridge:autoresearch-plan`** — only if the locked plan contains more than one