| name | dev |
| description | Pick the next unchecked task from the sprint and implement it with TDD. Reads board, posts available, implements, commits. |
| argument-hint | [sprint e.g. v1] |
| allowed-tools | Read, Write, Edit, Glob, Grep, Bash, mcp__ide__getDiagnostics |
You are implementing tasks for sprint "$ARGUMENTS" of this project.
Startup
- Read
kapi/board.md — check for blockers or decisions that affect your work
- Read
kapi/sprints/$ARGUMENTS/tasks.md — find the first unchecked task
- Read
kapi/sprints/$ARGUMENTS/prd.md — understand the sprint goal
- Post agent init to board.md:
- Announce: "Starting {T0N}: {title}"
TDD Cycle (strict — do not skip steps)
For each task:
- Read — understand what needs to be built (Files, Logic, Test fields in tasks.md)
- Implement — write the code. For skill files: write the SKILL.md with full instructions
- Verify — run the Test specified in the task:
- Build check:
npm run build (must pass)
- Type check:
npx tsc --noEmit (must pass)
- Manual verification as described in the task's
Test: field
- Mark done — change
- [ ] to - [x] in tasks.md
- Commit:
git add -A
git commit -m "feat(sprint-$ARGUMENTS): {T0N} - {short description}"
- Move to next unchecked task
When Blocked
- Add a note:
**BLOCKED:** {reason} under the task in tasks.md
- Write a blocker entry:
kapi/entries/{datestamp}-dev-blocker-{slug}.md
- Add to
## Active Blockers in board.md: - **Dev** — {reason} → entries/{filename}
- Skip to next task with no unresolved dependencies
When All Tasks Done
- Run
npm run build — must succeed
- Run
npx tsc --noEmit — must succeed
- Update
## Agent Status in board.md:
- **Dev** ($ARGUMENTS) — idle {ts}, all tasks committed, run /test
- Tell the user: "All tasks done. Run /test $ARGUMENTS to run the QA gate."
Conventions
- Skill files go in
.claude/skills/{name}/SKILL.md
- Entry files go in
kapi/entries/{YYYY-MM-DD-HHMM}-{role}-{type}-{slug}.md
- Never skip a task silently — always note blockers
- One commit per task