| name | planning-structure |
| description | Planning-track selection, epic/feature decomposition, readiness gates, and plan-delta rules. |
| user-invocable | false |
Planning Structure
Use this skill when producing or evaluating non-trivial implementation plans.
1. Track Selection
Choose the smallest track that still manages risk:
- Quick Change
- one localized change
- clear owner and file scope
- low ambiguity
- little or no architecture impact
- Feature Track
- one feature or bug spanning a few areas
- moderate ambiguity or dependencies
- clear user value and verification path
- System Track
- multi-surface work
- architecture, workflow, data-model, or integration impact
- cross-team or parallelization candidate
- requires explicit artifact planning or a readiness gate
Default upward when scope uncertainty is high.
2. Planning Shapes
Quick Change
Keep the plan compact:
- objective
- affected files and owners
- ordered steps
- verification
- scope boundaries
Feature Track
Use a feature-oriented structure:
- objective
- scope
- feature slices
- dependencies
- risks
- verification
- gaps and proposed defaults
System Track
Use a layered structure:
- objective
- scope and exclusions
- epics
- features inside each epic
- documentation artifacts needed
- dependencies and sequencing
- risks and open gaps
- readiness gate
- verification
3. Epic and Feature Rules
For larger work:
- Prefer
2-5 epics.
- Each epic should represent one coherent value stream or subsystem boundary.
- Each epic should contain a small number of implementable features or vertical slices.
- Each feature should state:
- user or system value
- deliverable
- verification signal
- main dependencies
Do not create epics for trivial work.
4. Deliverable and Verification Discipline
Every feature or major step should answer:
- what will exist when this is done?
- how will we verify it?
- what must already exist first?
Prefer concrete verification:
- tests
- commands
- observable UI flows
- logs or metrics
- API responses
5. Gaps and Proposed Defaults
When a decision is unresolved but planning can continue, include:
- Open Question
- Recommended Default
- Impact if wrong
If the unresolved point changes architecture, security, or data correctness materially, stop and clarify instead of assuming.
6. Documentation Artifacts
Only plan artifacts that materially reduce execution risk.
Common candidates for System Track:
architecture.md
data-model.md
api-contracts.md
runbook.md
privacy-security-notes.md
Do not auto-generate all of them. Name only what the work actually needs.
7. Readiness Gate
Execution is ready only when all are true:
- scope is stable enough to implement
- owners and affected areas are known
- dependencies are identified
- verification is concrete
- critical gaps are resolved or intentionally defaulted
Output:
Implementation Readiness: PASS
- or
Implementation Readiness: BLOCKED
If blocked, say exactly what must be resolved first.
8. Plan Delta
When scope changes after a plan already exists, do not rewrite from scratch first. Produce a delta:
- what changed
- what stays valid
- what steps are removed
- what new steps are added
- whether readiness changed
- whether routing or parallelization strategy changed
9. Clarification Bias
Treat planning as an interview loop, not a one-shot report.
Use clarification aggressively when the unresolved point affects:
- user-visible behavior, UX copy, or product semantics
- API contracts, schema, persistence, migration, or compatibility behavior
- security, privacy, performance, or reliability expectations
- verification, rollout, fallback, or non-goal boundaries
Default clarification behavior by track:
Quick Change: ask only when scope or acceptance criteria are not clear
Feature Track: expect at least one clarification round unless the request is already explicit
System Track: expect clarification and do not force execution-readiness until user-owned decisions are resolved
Gaps and Proposed Defaults is only for low-impact implementation details. If the gap changes UX, API, data correctness, security, or verification, ask instead of defaulting.
10. Anti-Patterns
Avoid:
- giant flat task lists with no value grouping
- steps with no deliverable or no verification
- architecture artifacts for small/local changes
- hiding unresolved critical gaps inside assumptions
- re-planning from zero when a plan delta is enough