| name | quick |
| description | Run a single sddx task through the PLAN → RED → GREEN → REFACTOR → VERIFY loop. Use when the user asks to execute one small, well-scoped task with spec-driven TDD. |
/sddx:quick
CLI: "${CLAUDE_PLUGIN_ROOT}/bin/sddx-run" "${CLAUDE_PLUGIN_ROOT}/dist/cli.mjs" (run from the repo root).
No task yet? Follow /sddx:plan first — task create --workspace branch
registers the spec and
switches to the sddx/<id> branch.
Run ... config show --output json once and keep .data.agent_model around
(the resolved config lives under data, not at the top level): this loop
runs solo (no subagent dispatch), so it has no per-role model to override —
agent_model only matters if a step here ever hands off to a dispatched
agent (e.g. via /sddx:plan's planner). Advisory only, not hook-enforced.
(--json still works as a deprecated alias for --output json, reading the
same .data.* shape.)
Then loop, recording every phase in the task file:
- RED — write the failing test FIRST. No implementation code of any kind.
Run the test runner and capture its exit code, then:
... task phase <id> RED --test-exit <code>
(rejected unless the code is nonzero — a passing test is not RED).
- Red-check —
... red-check <id> runs the spec's oracle and must see it
fail. Exit 0 means the oracle cannot discriminate — fix the spec before any
implementation; verify refuses a task without this record.
- GREEN — write the minimal implementation. Re-run the tests; when they
pass:
... task phase <id> GREEN --test-exit 0.
- REFACTOR (optional) — clean up; tests must stay green.
... task phase <id> VERIFY, then follow /sddx:verify.
Rules:
- Phase transitions demand evidence (test exit codes); never claim a phase.
- Check
iterations in .sddx/tasks/<id>.json against the spec's
max_iterations (default 5) each verify attempt; exceeded → stop, escalate.
- Never touch
.sddx/receipts/**. DONE is the verifier's call, not yours.
- After DONE (or whenever the loop pauses for input), run
... next-actions
and relay its output verbatim as the completion message — don't compose
your own summary of what's next. On the user's reply, run
... next-actions --select "<reply>" and relay that output too. Cleanup:
... cleanup <id> (deletes only merged branches).
