| name | tdk-task-breakdown |
| description | Generate parent epic child-spec-seed breakdown artifacts from epic PRD plus /tdk-epic-hld context. Use before child /tdk-specify loops. |
| metadata | {"version":"1.0.0"} |
tdk-task-breakdown
Create parent epic child-spec-seed Markdown from epic PRD and epic HLD context.
User Input
$ARGUMENTS
Trigger: /tdk-task-breakdown <epic-id> [--force]
Boundary Declaration
This command produces:
- Epic dashboard update at
{FEATURE_DIR}/index.md
- Task-breakdown stage manifest at
{FEATURE_DIR}/tasks-breakdown.md
- Markdown child spec seed artifacts under
{FEATURE_DIR}/tasks-breakdown/
tasks-breakdown/task-NNN-{slice}.md files
This command does NOT:
- Create child
spec.md files (use /tdk-specify <child-id> "<seed>")
- Clarify child specs (use child
/tdk-clarify)
- Create implementation plans (use child
/tdk-plan)
- Implement code (use child
/tdk-implement)
- Create GitHub, GitLab, Backlog, or other tracker issues
- Call external tracker APIs or CLIs
- Mint
UR-*, FR-*, SC-*, or FS-* identifiers
Core output is tracker-neutral child spec seed Markdown. Child specs are the
implementation units after this stage.
Skill References
Load before deriving any breakdown content:
references/task-breakdown-output-contract.md
Execution Steps
Step 0 - Validate Epic ID
Invoke tdk-validate-task-id with $ARGUMENTS and host skill name /tdk-task-breakdown.
If STOP, halt execution.
Store: TASK_ID, TASK_ID_SOURCE.
Step 0.1 - Load Project Context
Invoke tdk-load-project-context with validated TASK_ID.
Store: PROJECT_CONTEXT, FEATURE_DIR.
Step 1 - Read Epic PRD
Require and read:
{FEATURE_DIR}/epic-prd.md
{FEATURE_DIR}/epic-prd/prd.md
{FEATURE_DIR}/epic-prd/slice-map.md
{FEATURE_DIR}/epic-prd/open-questions.md
If any file is missing, STOP before writing and tell the user to run
/tdk-epic-prd {TASK_ID} first.
Step 1.5 - Require Epic HLD Context
Require {FEATURE_DIR}/high-level-design.md and read the artifacts it
lists. If it is missing, STOP and tell the user to run /tdk-epic-hld {TASK_ID}
first.
HLD is parent design context only: it sharpens boundaries, dependencies,
assumptions, risks, and child spec seed wording. It is not requirement authority.
Step 2 - Parent Readiness Gates
STOP before writing any file when:
epic-prd/open-questions.md has any item under ## Blocking Questions.epic-prd/slice-map.md has no independently specifiable slice.epic-prd/slice-map.md contains catch-all slices such as "all features",
"entire MVP", or "whole epic".high-level-design.md does not mark the HLD set ready for task breakdown.
Step 3 - Load Output Contract
Read references/task-breakdown-output-contract.md from this skill directory.
Use that reference as the single source of truth for:
- Output filenames
- Frontmatter schema
- Required sections
- Slice granularity rules
- Child spec seed rules
- Requirement-authority boundaries
Step 4 - Extract Source Slices
From epic-prd/slice-map.md, extract slice keys, capability, actor, outcome,
dependencies, suggested child spec title, priority, and seed text.
From HLD artifacts, extract only boundary, dependency, data/user flow,
interface, assumption, risk, and follow-up context that affects child spec seed
quality.
Do not invent file paths, APIs, database tables, owners, estimates, tracker
labels, or formal requirement IDs unless the epic PRD or HLD explicitly states
them as assumptions to validate.
Step 5 - Derive Child Spec Seeds
Create one breakdown file for each independently specifiable slice:
- Each item must cite a source slice key.
- Each item must include source PRD/HLD references.
- Each item must include suggested
/tdk-specify <child-id> "<seed>" text.
- Each item must define boundary, dependencies, assumptions/risks, and what to
clarify in the child spec.
- Prefer fewer coherent child specs over fragmented one-line seeds.
If a slice is too broad to become a child spec, STOP and tell the user to refine
the epic PRD slice map before writing output.
Step 6 - Write Markdown Artifacts
If tasks-breakdown/index.md exists and sibling tasks-breakdown.md is missing,
STOP with:
legacy layout detected: tasks-breakdown/index.md is from the old nested-manifest layout. Re-run with --force to regenerate task breakdown in the new layout, or recreate the test epic. TDK does not auto-migrate old tasks-breakdown/index.md content.
Create or update only:
{FEATURE_DIR}/tasks-breakdown.md{FEATURE_DIR}/tasks-breakdown/task-NNN-{slice}.md
Do not write tasks.md, child spec.md, plan files, tracker config, or
implementation files.
tasks-breakdown.md is the authoritative manifest for the current generated
set. Consumers and agents must read the child spec seed files listed in the
stage manifest, not glob every file in tasks-breakdown/.
Step 6.5 - Update Epic Dashboard
Update the generated task-breakdown section in {FEATURE_DIR}/index.md. Link
the tasks-breakdown.md stage manifest, summarize seed count and tracker
boundary, and show the next command for child /tdk-specify. Preserve
user-owned content outside generated sections. Replace an existing generated
task-breakdown section only after confirmation or when --force is supplied.
Step 7 - Report Results
Report:
- Number of child spec seed files written
- Relative paths for
{FEATURE_DIR}/index.md, tasks-breakdown.md, and each seed file
- Reminder: child specs start with
/tdk-specify <child-id> "<seed>"
- Reminder: tracker issue creation is consumer-owned and out of TDK core scope
Step 8 - Recommend Next Step
Use AskUserQuestion with header "Next Step" after reporting results:
| Option | Action |
|---|
Start first child /tdk-specify (Recommended) | Read the first seed listed in tasks-breakdown.md, then invoke its suggested /tdk-specify <child-id> "<seed>" command exactly |
| Choose a different child seed | Stop and tell the user to choose one seed file from tasks-breakdown.md; do not infer a different child order |
| End session | Stop |
Quality Gates
