speckit-workflow

name	speckit-workflow
description	Teaches the Swarm coordinator to follow Speckit's pre-decomposed tasks.md as the authoritative task breakdown instead of generating its own decomposition via CASS.
tags	["speckit","workflow","decomposition"]

Speckit Workflow — Swarm Skill

This skill teaches you (the Swarm coordinator) how to work with Speckit's pre-decomposed task lists. When a tasks.md file exists for the active feature, it is the authoritative decomposition — do NOT re-decompose via CASS.

When This Skill Applies

Check for a tasks.md file in the active spec directory:

1. Look for specs/NNN-*/tasks.md relative to the project root (Speckit strategic workflow)
2. Look for openspec/changes/*/tasks.md (OpenSpec tactical workflow)

If tasks.md exists, follow the instructions below. If no tasks.md exists, proceed with standard CASS decomposition.

Pre-conditions

CRITICAL: All work MUST be committed and pushed on the current feature branch before any branch switch occurs.

• After completing all phases, commit and push all changes before suggesting PR creation, merging, or switching to main.

• Before creating a new feature branch (via /speckit.specify), check git status --short for uncommitted changes. If uncommitted changes exist, use the question tool to confirm before proceeding. Include the git status --short output in the question so the user can see which files are uncommitted:

  "Uncommitted changes detected. Switching branches with a dirty working tree may cause changes to be applied to the wrong branch or lost."

  Use options: ["Stash changes and continue", "Abort -- keep changes as-is"]

  • If the user selects "Stash changes and continue": run git stash. If git stash returns a non-zero exit code, STOP the workflow, report the stash failure to the user, and do NOT proceed to branch creation. If git stash succeeds, re-run git status --short to verify the working tree is clean. If the working tree is still not clean, STOP and report the remaining uncommitted changes. Only when the working tree is confirmed clean, proceed to branch creation. Upon workflow completion, inform the user that their changes are stashed and can be restored with git stash pop.
  • If the user selects "Abort -- keep changes as-is": STOP the workflow immediately without creating a branch or modifying the working tree.

• Never silently switch branches with a dirty working tree. Uncommitted changes may follow to the wrong branch or be lost entirely.

Reading tasks.md

Phase Structure

Tasks are organized into numbered phases:

## Phase 1: Setup
## Phase 2: Foundational
## Phase 3: User Story 1 (P1)
## Phase 4: User Story 2 (P1)
...
## Phase N: Polish

Map each phase to a Swarm epic. Phase 1 becomes epic "Setup", Phase 2 becomes epic "Foundational", etc.

Task Format

Each task follows this format:

- [ ] T001 [P] [US1] Description with file path

• Checkbox - [ ]: Uncompleted task. - [x]: Done.
• Task ID T001: Sequential identifier
• [P] marker: This task can run in parallel with other [P] tasks in the same phase (they touch different files). Tasks WITHOUT [P] are sequential.
• [US1] label: Maps to User Story 1 from the spec. Use as cell metadata when creating Swarm cells.
• Description: What to do and which file to modify.

Phase Dependencies

• Setup and Foundational phases MUST complete before any User Story phases begin.
• User Story phases may run in parallel if they are marked as independent in the Dependencies section of tasks.md.
• Polish phase runs after all User Story phases.

Creating Swarm Work Items

1. Read the full tasks.md to understand all phases.

2. Create one epic per phase using hive_create_epic():

   • Epic title = phase name (e.g., "Phase 3: US1 - Doctor Core")
   • Subtasks = the - [ ] tasks within that phase

3. For tasks marked [P]: These can be spawned as parallel workers. Each [P] task touches different files, so file reservation conflicts are unlikely.

4. For tasks without [P]: These MUST be executed sequentially within the phase. Spawn one worker at a time, waiting for completion before the next.

5. For tasks already checked [x]: Skip these — they are already completed.

Worker Instructions

When spawning workers via swarm_spawn_subtask():

1. Include the task description and file path in the worker's prompt
2. Workers MUST call swarmmail_reserve() on the target file(s) before editing
3. Workers MUST mark the task [x] in tasks.md after completing it
4. Workers MUST call swarm_complete() when done

Phase Checkpoints

After all tasks in a phase are complete:

1. Run the project's test suite (check AGENTS.md for the build/test commands)
2. If tests fail, create a new cell to fix the failure before advancing to the next phase
3. Report phase completion via swarm_status()

Prerequisite Skill

Load unbound-force-heroes alongside this skill to get hero routing and workflow stage context:

skills_use({ name: "unbound-force-heroes" })
skills_use({ name: "speckit-workflow" })

Entry Point

The /uf.unleash command is the primary way to trigger autonomous pipeline execution using this skill. It orchestrates the full Speckit pipeline (clarify, plan, tasks, spec review, implement, code review, retrospective, demo) and uses the task format described above for the implementation phase. /uf.unleash also supports OpenSpec (opsx/*) branches.