GStack Task Planner
Use this skill as a portable planning workflow for coding agents.
This skill is designed for general agent environments. It does not assume a specific project-management tool, planning system, or host runtime. Adapt the workflow to the tools available in the current environment.
Its job is to turn an ambiguous or broad request into a concrete plan:
- Clarify the real goal
- Identify scope and constraints
- Break the work into executable pieces
- Surface assumptions and risks early
- Define a verification strategy before implementation
- Stop at the planning stage unless the user explicitly asks to proceed
Purpose
This skill helps an agent turn a request into a clear implementation plan:
- Understand the actual task and success criteria
- Gather the minimum useful context
- Break the work into coherent subproblems
- Identify risks, unknowns, and dependencies
- Define validation and rollout checks
- Deliver a practical plan that another agent or engineer can execute
When To Invoke
Invoke this skill when:
- The user asks to plan a feature or task
- The user wants a work breakdown before coding
- The request is broad, ambiguous, or multi-step
- The user wants implementation steps, scope, or milestones
- The agent should pause and think before building
Do not invoke this skill when:
- The user clearly wants immediate implementation and the task is trivial
- The work is already fully scoped and only execution remains
- The user wants a bug investigation rather than a forward plan
Operating Principles
- Prefer clarifying the goal before proposing solutions
- Prefer plans that are executable, not aspirational
- Prefer smaller, independently verifiable steps
- Prefer explicit assumptions over hidden ones
- Prefer identifying uncertainty early rather than during implementation
- Do not expand scope just because more work is possible
Workflow
0. Confirm The Request
Restate the request in concrete terms:
- What the user wants changed or built
- Why it matters
- What success likely looks like
If the task is underspecified, ask only the most important clarifying questions.
1. Gather Planning Context
Collect just enough context to make a useful plan:
- Read the relevant files, docs, or architecture notes
- Identify the likely feature area or system boundary
- Check existing conventions and adjacent implementations
- Identify technical or product constraints
Do not do a full implementation investigation unless needed for planning.
2. Define Scope
Separate the work into:
- In scope
- Out of scope
- Assumptions
- Unknowns that may block execution
If the request is too large, propose a narrower first slice.
3. Break Down The Work
Produce a step-by-step plan with meaningful boundaries.
Good plan steps are:
- Small enough to execute
- Large enough to represent a coherent unit of work
- Ordered by dependency
- Easy to verify independently
Prefer plans that distinguish:
- Design or schema changes
- Backend logic changes
- Frontend or API surface changes
- Tests and verification
- Documentation or rollout tasks
4. Identify Risks And Dependencies
For each meaningful part of the plan, consider:
- Dependency on another module, service, or team
- Migration or compatibility risk
- Data-shape or state-transition risk
- Auth, permission, or trust-boundary risk
- Performance or scale risk
- UX or user-flow regression risk
Do not overdo this section. Focus on risks that change execution decisions.
5. Define Verification
Every plan should include a validation path before coding starts.
Include the strongest likely checks such as:
- Unit or integration tests
- Typecheck or lint
- Build or compile validation
- Manual flow checks
- Monitoring or rollout verification
If verification is unclear, call that out as a planning gap.
6. Deliver The Plan
End with a concrete output that can drive execution.
The final plan should usually include:
- Goal
- Scope
- Assumptions
- Step-by-step plan
- Risks
- Verification approach
Unless the user explicitly asks to continue, stop after presenting the plan.
Tooling Guidance
Adapt to the current host's tools.
If file search and file read tools exist:
- Use them to understand the current system before planning changes
If issue trackers, tickets, or design docs exist:
- Use them to reconcile the request with prior decisions
If architecture or test docs exist:
- Use them to identify likely integration points and validation steps
If a tool does not exist in the current host:
- Degrade gracefully and state which planning assumptions remain unverified
Output Contract
For planning tasks, the final answer should usually include:
- Goal
- Scope
- Assumptions or open questions
- Step-by-step implementation plan
- Risks
- Verification plan
For larger tasks, also include:
- Recommended first slice
- Dependencies or sequencing notes
If the plan is blocked by missing information, say exactly what is missing.
Anti-Patterns
Do not:
- Jump into implementation without separating planning from execution
- Create a plan that is broader than the request
- Produce steps that cannot be independently verified
- Hide major assumptions
- Confuse architecture speculation with a practical execution plan
- Pretend the plan is final when key information is still missing
Example Invocation Scenarios
- "Break this feature into steps"
- "Help me plan this implementation"
- "Decompose this task before we code"
- "What is the execution plan for this change?"
- "Turn this requirement into an actionable plan"