sprint
Feature Forge — preview
--farm — preview. This flag is part of the Feature Forge and ships dormant by default. Promotion to stable is driven by real-world evidence.
Autonomous sprint — one interactive spec gate, then plan-to-PR execution with every auto-decision SMARTS-scored and logged. Hard gates remain true stops.
What it does
Section titled “What it does”This is codeArbiter’s autonomy mode. It collapses the usual spec-plan-execute cycle into one interactive checkpoint up front, then runs the rest of the arc — task selection, test-first implementation, review, and landing the branch — without stopping between batches. Every call it makes on your behalf along the way is scored and appended to a running log with a confidence rating, so a low-confidence choice is easy to find and revisit later even though nobody was there to approve it in the moment.
/ca:sprint [goal] [--farm]A short goal seeds the spec conversation that opens the sprint. Add --farm to send
implementation work to lower-cost worker agents while the same spec-authoring, test-writing, and
review responsibilities stay put.
Example
Section titled “Example”> /ca:sprint add rate limiting to the public webhook endpoint
Opening the sprint spec conversation...[Socratic refinement — per-IP limit, window, response on exceed]Spec written to .codearbiter/specs/webhook-rate-limit.md. Plan written with 5 tasks.Approve spec and plan to begin autonomous execution? (y/n)> y
Executing task 1/5... task 2/5... task 3/5 touches security-controls.md scope — this is a hard gate,halting for your decision rather than choosing on your behalf.
Task 3 adds a Redis-backed limiter key derived from client IP. Confirm this fits the approvedsecurity posture before I continue? (y/n)> y
...task 4/5... task 5/5 complete. Every auto-decision logged to sprint-log.md (1 low-confidence entry).Routing to commit-gate, then finishing-a-development-branch (opening a PR — sprint never merges).When to reach for it
Section titled “When to reach for it”Reach for /ca:sprint when the spec can be made concrete up front and you want the whole plan to run
without per-batch checkpoints. /ca:feature is the better fit when you want to review progress
between batches.
| Gate | When | Effect |
|---|---|---|
| spec + plan approval | before autonomous execution starts | hard stop; both the sprint spec and the resulting plan need your explicit sign-off |
| hard gates | a security boundary, crypto/secret handling, an irreversible operation, an override request, an unresolved confirmation point, or a merge to the default branch comes up mid-run | execution halts and surfaces the decision to you instead of choosing on your behalf |
Related
Section titled “Related”Source
Section titled “Source”Source — plugins/ca/commands/sprint.md (v2.9.1)
---description: Autonomous sprint — one interactive spec gate, then plan-to-PR execution with every auto-decision SMARTS-scored and logged. Hard gates remain true stops.argument-hint: "[goal] [--farm]"---
# /ca:sprint — autonomous sprint
The autonomy mode. Brainstorm a sprint spec with the user — the one interactive gate — then executethe approved plan end-to-end without per-batch checkpoints, deciding "as the user" via SMARTS onevery non-hard-gate point. Every auto-decision lands in `.codearbiter/sprint-log.md` (append-only)with a confidence flag; the `low`-confidence entries are exactly what the user reviews afterward.Nothing is hidden behind autonomy.
## Flow
Load and follow `${CLAUDE_PLUGIN_ROOT}/SPRINT.md` — it is the procedure. In brief:
1. **Sprint spec (STOP)** — `brainstorming` scoped to a sprint, then `writing-plans`. Explicit user approval of spec AND plan before autonomy begins.2. **Autonomous execution (BLOCK)** — `subagent-driven-development` runs the plan; test-first via `tdd`, two-pass reviewed, fresh-run verified. SMARTS decides non-hard-gate points; everything logs.3. **Land & summarize (BLOCK)** — `commit-gate`, then `finishing-a-development-branch`, which auto-selects open-PR. `/ca:sprint` never merges and never discards; the merge decision is the user's.
Hard gates — `security-controls`, crypto/secrets/auth, irreversible ops, `/override`, anunresolvable `[CONFIRM-NN]`, merge-to-default — are NEVER auto-decided. They halt and surface.
## Arguments
- **`"goal"`** — seed for the sprint-spec brainstorm.- **`--farm`** — cost-arbitrage backend: cheap workers implement under the same gates; Claude still authors spec, failing tests, plan, and reviews everything. Pre-flights `FARM_API_KEY`.
## Routes to
`${CLAUDE_PLUGIN_ROOT}/SPRINT.md` (mode body), which routes through `brainstorming`,`writing-plans`, `subagent-driven-development`, `commit-gate`, `finishing-a-development-branch`.
## When NOT to use
- A single feature with human checkpoints → `/ca:feature`.- Work whose spec cannot be made concrete up front — the one interactive gate is load-bearing; a thin spec makes hard-gate stops frequent instead of rare.