← ClaudeAtlas

buildlisted

Implement the next unblocked phase of a plan with feedback loops, mark its checkboxes, offer a commit. Use after /plan when `.specs/plans/<slug>.md` exists, or when asked to build the next phase. Don't use for unplanned changes (/tdd) or to check progress (/test).
helderberto/agent-skills · ★ 16 · Web & Frontend · score 81
Install: claude install-skill helderberto/agent-skills
# Build Phase Implement the next incomplete phase of a plan — one phase per invocation. ## Input `$ARGUMENTS` is a `<slug>` or an `@path` (read that file directly as the plan). If empty or the plan is missing, list `.specs/plans/*.md` as numbered options and wait. ## Workflow ### 1. Find the next incomplete phase Scan `## Phase N` headings and count `- [ ]` / `- [x]`. The next incomplete phase is the first with an unchecked item **and** whose `**Blocked by**` phases (if declared) are all complete. If the first incomplete phase is blocked, pick the next unblocked one and say why. All phases complete → "All phases complete. Run `/test <slug>` to verify." Stop. ### 2. Present the phase Show the title and unchecked items. If on the default branch, offer to create `feat/<slug>` (Recommended) before touching code. ### 3. Implement For each unchecked item, in order: read the plan's architectural decisions and the item's context, explore the surrounding code, implement following the project's conventions (CLAUDE.md, linter, test setup), write tests alongside. Stay inside the phase boundary — never implement items from other phases. Never impose conventions the project doesn't already use. ### 4. Feedback loops Detect and run the project's checks (types, tests, lint, format) — prefer tasks the project defines; detection table in [validate-code](../validate-code/SKILL.md). Fix and re-run until green. A check still failing after 3 attempts is a **blocker**: report the last