Install Bmad Handoff
Pick your agent. Every command is ready to paste.
npx skills add aj-geddes/claude-code-bmad-skills --skill bmad-handoffThe open skills CLI detects your agents (Claude Code, Cursor, Codex and 70+ more) and installs to the right folder. Add -g to install globally.
01
Install it
Run the command above, or upload the folder in Claude.
02
Ask naturally
Describe the task. Your agent loads bmad-handoff when the request matches its description.
03
Name it if needed
If it does not trigger, say “use the bmad-handoff skill”. In Codex, type $bmad-handoff.
BMAD Handoff
Purpose: Scan the planning output folder for all stories at status ready-for-dev,
compile them into a single handoff-manifest.json, and leave it where any downstream
dev tool can read it — without coupling to any specific runner.
This is the last artifact the planning plugin produces. What happens next is owned by the external dev tool, not by this skill.
When to Run
Run this skill after the story-writing phase is complete and you (or the user) have
confirmed that at least one story carries status ready-for-dev. The manifest is a
point-in-time snapshot; re-run the skill to refresh it.
Workflow
Use TodoWrite to track progress through these steps.
1. Locate the output folder
Look for bmad-output/project-context.md (the default output folder) or ask the
user for the output folder. Default: bmad-output/.
2. Discover story files
Glob for **/{epic}.{story}.*.story.md under the output folder.
Accept alternative flat layouts (stories/*.story.md) if the glob turns up nothing.
3. Filter to ready-for-dev
Read the **Status:** header field of each story file (in the story header block, not a
## Status heading). Include only stories whose status value is exactly ready-for-dev.
If none are found, report which statuses were seen and stop — do not produce an empty manifest.
4. Extract per-story fields
For each qualifying story file, extract:
| Field | Source in story file |
|---|---|
id |
Filename stem or ## Story heading ID |
storyFilePath |
Relative path from project root |
status |
**Status:** header field value |
epic |
First segment of filename, e.g. "2" in 2.1.stripe.story.md |
storyNumber |
Second segment, e.g. "1" |
title |
First H1 or ## Story heading text |
ownedScope |
## Owned File/Module Scope — list every path verbatim |
wave |
## Dependency Maps → wave/parallel_set annotation (integer or null) |
parallelSet |
Same section — parallel set label if present (string or null) |
dependencies |
## Dependency Maps → blocked-by story IDs (array, may be empty) |
acceptanceCriteriaSummary |
First 3 AC items from ## Acceptance Criteria, each ≤120 chars |
lockedSectionsNote |
Static string — see schema |
devAgentRecord |
Static null — placeholder for the dev tool to populate |
5. Compute wave order
If stories do not already carry explicit wave annotations:
- Stories with empty
dependenciesarrays are wave 1. - A story whose every dependency is in wave N or lower is wave N+1.
- Add the computed
wavevalue; leaveparallelSetnull when not annotated.
6. Write the manifest
Write to {outputFolder}/handoff-manifest.json.
Use the schema from
${CLAUDE_PLUGIN_ROOT}/skills/bmad-handoff/templates/handoff-manifest.schema.json
as the structural contract. Populate schemaVersion: "1.0".
Sort stories by wave ascending, then by id ascending within each wave.
7. Report to the user
Print a compact summary:
Handoff manifest written → bmad-output/handoff-manifest.json
schemaVersion : 1.0
stories : <N> ready-for-dev
waves : <W> (wave 1 has <X> stories, can start immediately)
output path : bmad-output/handoff-manifest.json
If any story was missing required sections (e.g. no ## Owned File/Module Scope),
list those as warnings — do not silently omit or fabricate data.
Manifest Field Definitions (Quick Reference)
See REFERENCE.md for the full schema narrative and adapter notes.
| Field | Type | Required | Notes |
|---|---|---|---|
schemaVersion |
string | yes | Semver string; current = "1.0" |
generatedAt |
string | yes | ISO-8601 UTC timestamp |
projectName |
string | yes | From project-context.md or user input |
outputFolder |
string | yes | Relative path used to find stories |
stories |
array | yes | One object per ready-for-dev story |
stories[].id |
string | yes | Unique story identifier |
stories[].storyFilePath |
string | yes | Relative path to the .story.md file |
stories[].status |
string | yes | Always "ready-for-dev" in this manifest |
stories[].epic |
string | yes | Epic identifier |
stories[].storyNumber |
string | yes | Story number within epic |
stories[].title |
string | yes | Human-readable story title |
stories[].ownedScope |
array | yes | File/module paths this story may modify |
stories[].wave |
integer | yes | Execution wave (1 = no dependencies) |
stories[].parallelSet |
string|null | no | Label if explicitly grouped |
stories[].dependencies |
array | yes | Story IDs that must complete first |
stories[].acceptanceCriteriaSummary |
array | yes | First 3 AC items, ≤120 chars each |
stories[].lockedSectionsNote |
string | yes | Instruction to dev tools |
stories[].devAgentRecord |
null | yes | Dev tool populates; always null at emit time |
lockedSectionsNote is always the string:
"Sections Acceptance Criteria, Dev Notes, and Testing are LOCKED. External dev tools must not edit them. Populate only the Dev Agent Record section."
Subagent Strategy
For small backlogs (≤15 stories), this skill runs single-threaded — the extraction loop is fast and context fits in one session.
For large backlogs (16+ stories), fan out story extraction in parallel:
| Agent | Task |
|---|---|
| Agent 1…N | Read story files in their assigned slice; extract fields; return JSON fragment |
| Coordinator | Merge fragments; compute wave order; write manifest |
Each agent receives the list of file paths for its slice plus the field extraction table above. It returns a JSON array of story objects (no wave field yet). The coordinator merges, computes waves, sorts, and writes the manifest.
Key Guidelines
- Never fabricate field values — if a field is missing from a story file, emit
nulland add a warning to the summary. - Do not modify story files — this skill is read-only with respect to story content.
ownedScopeis critical for parallel-conflict safety; never collapse or summarize it.- This manifest is a STABLE versioned interface. Increment
schemaVersion(as a separate schema revision, not within this run) before adding or removing fields. - Do not filter out stories based on anything other than
ready-for-devstatus — dependency resolution is the dev tool's job. - If the user asks to "re-run" or "refresh" the manifest, overwrite the existing file.
Part of the BMAD Planning & Orchestrator plugin — a Claude Code harness for the BMAD Method by the BMAD Code Organization (https://github.com/bmad-code-org/BMAD-METHOD). Implements the spirit of
bmad-handoff. All methodology credit belongs to the BMAD Code Organization.
Tips for Bmad Handoff
- • Start a fresh session after installing so your agent picks up the new skill.
- • It bundles reference files. The agent reads them only when needed, so they cost no context until then.
- • It is part of a pack with 20 skills. Install only the ones you need with --skill to keep triggers precise.
- • If two skills overlap, mention the one you want by name.
Questions about Bmad Handoff
What does the Bmad Handoff skill do?
Emits a dev-tool-agnostic handoff manifest from ready-for-dev stories so an external dev plugin or runner can pick up and execute the work. Use when the user says "generate a handoff", "create handoff manifest", "export stories for dev", "hand off to dev tool", "produce handoff manifest", "ready to hand off", "prepare handoff for external tool", "export ready-for-dev stories", or "create the handoff package". Also tr
How do I install Bmad Handoff?
Run "npx skills add aj-geddes/claude-code-bmad-skills --skill bmad-handoff" in your project. The skills CLI installs it for Claude Code, Cursor, Codex and 70+ other agents. In Claude on the web, upload the skill folder as a ZIP under Customize > Skills.
Does Bmad Handoff work in Cursor and Codex?
Yes. Bmad Handoff uses the open SKILL.md format, so it works in Claude Code, Claude, Cursor, OpenAI Codex, Gemini CLI, GitHub Copilot and other agents that support agent skills.
Is Bmad Handoff free?
Bmad Handoff is open source on GitHub. It is free to install; check the license before commercial use.
Who made Bmad Handoff?
Bmad Handoff is published by aj-geddes in the aj-geddes/claude-code-bmad-skills repository, together with 19 other skills.





