speckit-beads-implement
Implement dependency-ready Beads tasks with claim and close semantics
Files1
View on GitHubSKILL.md.specify/extensions/beads/.specify-dev/agent-commands/claude/speckit-beads-implement/SKILL.md
GitHubUser Input
$ARGUMENTS
Execute the current Spec Kit feature from Beads. Do not require or update tasks.md.
- Resolve the current
FEATURE_DIRwith.specify/scripts/bash/check-prerequisites.sh --json --require-specand derive the feature labelspeckit:<feature-directory-name>. - Inspect every file under
FEATURE_DIR/checklists/without modifying it. If any checklist contains unchecked items, report the counts and ask before proceeding. - Run
bd ready --json. Keep tasks labeled bothspeckitand the current feature label. - Select the highest-priority ready task; break ties by
metadata.speckit.display_id. If the user supplied a Beads ID, verify that task is ready instead. - Inspect it with
bd show <id> --json. Use its stored description, design context, acceptance criteria, notes, and source references. Inspect the parent epic for the stored specification and technical design; verify source artifacts against their hashes when the checkout is available. - Atomically claim it with
bd update <id> --claimbefore editing. - Implement only that task's scope, respecting its acceptance criteria and the technical plan. Run the relevant focused tests and quality gates.
- On success, close it with
bd close <id> --reason "Implemented and validated". On failure or a blocker, leave it open or in-progress, record the blocker or discovered task in Beads, and report it. - Show newly ready tasks with
bd ready. Continue only when the user's arguments requestall; otherwise stop after one task.
Rules:
- Never mark work complete solely because files changed; acceptance criteria and validation must pass.
- Create discovered follow-up work in Beads and wire its dependencies.
- Do not maintain a parallel Markdown checklist.
- Do not commit or push unless separately authorized.