speckit-implement

Imported from rayedbajwa/spaces (data/speckit/skills/speckit-implement/SKILL.md). Install upstream with npx skills add rayedbajwa/spaces --skill speckit-implement. Copyright stays with the author.

Implement: carry out the tasks

User input (consider it if not empty): $ARGUMENTS

Steps

1. Run .specify/scripts/bash/check-prerequisites.sh --json --require-tasks --include-tasks; use FEATURE_DIR and AVAILABLE_DOCS. If tasks.md is missing, say to run tasks first and stop.
2. Checklists: if FEATURE_DIR/checklists/ has files, count - [ ] vs - [x] per file. Open items do not block (the run's review gates decide): keep them in view while implementing, and list in your report the ones still open, with why.
3. Read tasks.md and plan.md; read data-model.md, contracts/, research.md, quickstart.md only when a task needs them.
4. Ignore files: for a git repository, make sure .gitignore (and, when the tool is in use, .dockerignore, .eslintignore/eslint ignores, .prettierignore, .npmignore when publishing, .terraformignore, .helmignore) covers the stack's build output, dependencies, logs, env files and secrets. Append missing patterns only; never remove lines.
5. Execute phase by phase, in task order:
   • Finish a phase before the next. [P] tasks may run together; tasks touching the same file run in sequence.
   • Tests before the code they cover, when the tasks include tests.
   • After each task, tick it in tasks.md (- [x] T012 …) and run the tests for what it changed.
   • A failing non-parallel task stops the phase: fix it, or report the error with context and what is needed.
6. Finish: every required task done and ticked, the implementation matches the spec and plan, and the tests you ran pass. Report what was completed and anything left.