| name | implement |
| description | Full autonomous feature implementation from customer spec to PR. Use when user provides a customer spec (file path, inline text, or ticket reference) and wants end-to-end implementation including spec/plan artifacts, feature branch, coding, verification, commits, and PR creation. |
| allowed-tools | Bash, Read, Write, Edit, Glob, Grep, Task, WebFetch, Skill |
| argument-hint | <path-to-spec | inline text> |
Context
- Today: !
date +%Y-%m-%d
- Time: !
date +%H-%M
- Branch: !
git branch --show-current
- Status: !
git status --short
Step 1 - Spec
Read $ARGUMENTS (file path or inline text). Generate a kebab-case slug from it.
Create .aiwork/{date}_{slug}/spec.md per the .aiwork protocol.
First, copy the FULL original text into a blockquote. This is a mechanical paste - never summarize, truncate, or rephrase.
Then write analysis sections: summary, decisions table (Decision | Choice | Rationale), scope (in/out), acceptance criteria. Mark ambiguous points [DECIDED].
Step 2 - Branch & plan
Create feature branch. Explore codebase for similar patterns and conventions. Use MCP tools for external services if available.
For complex tasks, write plan.md next to spec with numbered implementation steps. Mark unclear items [OPEN].
Step 3 - Implement
Execute plan. Use Task agents for parallel independent work.
Step 4 - Improve
Improve all rendered text with /czech-typography skill.
Step 5 - Verify
Visual check in the browser.
Step 6 - Ship
Commit changes (split if logically separate).
Ask whether to create a PR.
If yes: Create PR with gh pr create (summary, changes, acceptance criteria). Push.
Output report: branch, PR URL, files changed, verification status, open questions.
