do-sdlclisted
Install: claude install-skill tomcounsell/ai
# do-sdlc — SDLC Pipeline Supervisor and Single-Stage Router
This skill is the **one substantive SDLC skill**. It has two entry modes over one shared step
spine:
| Mode | Entry | Executes | Contract |
|------|-------|----------|----------|
| **Router mode** | a single-stage router invocation | Steps 1–4 once | dispatch ONE sub-skill, then return; the supervising session re-invokes |
| **Supervisor mode** | `/do-sdlc` | Steps 1–7 | loop until merge/blocked |
Router mode assesses state and dispatches exactly one `/do-*` skill, expecting the supervising
session to re-invoke it after each stage completes. Supervisor mode is the local stand-in for that
session: it re-invokes the router itself, dispatching each stage to a subagent on the
stage-appropriate model (opus/sonnet), until merge, a blocking guard, or the iteration cap. You
are the supervisor or router, never the worker: you assess, dispatch, and track, and the stage
subagents do all the work.
**Redundant-context check.** If a live supervising context already owns this issue, supervisor
mode is redundant — that context IS the supervision loop. Do not run it; drive the pipeline one
stage at a time through router mode instead. A live *supervised-run signal* for the issue number
is a different case: if its `owner_run_id` is one this run already holds, it is this run's own
hand-off and must be inherited, not treated as grounds to stand down (see Step 2).
## Repo Context Probe
If `docs/sdlc/do-sdlc.md` exists, read it and