strategic-planning-protocol

Strategic Planning Protocol (SPP-001)

Use this skill to transform a draft plan into a comprehensive, persona-validated roadmap with identified gaps, required artifacts, and success-enabling pre-work.

When to Use

• Before executing any multi-phase project (3+ phases)
• After creating an initial plan but before implementation
• When inheriting or reviewing an existing roadmap
• When a project spans multiple packages, teams, or domains
• When you need to ensure alignment across different stakeholders

Core Principle

Plans fail not from bad ideas but from missing foundations. This skill ensures you identify and address gaps before they become blockers.

---

Phase 0: External Knowledge Assessment

Before evaluating the plan, assess what external knowledge is needed.

0.0 Determinism + Registry Hygiene (Monorepo Reality Check)

If the plan adds or changes any Capability/Blueprint/Tool that is expected to be discoverable (MCP/UI/registry), explicitly include these checks up front:

• Deterministic artifacts are part of the contract: some repos require committed/generated artifacts (ex: tool catalogs) to be updated alongside code changes.
• Registry / catalog update order matters (avoid “tool exists in code but is missing in UI”):
  • Align metadata.domain/subdomain with metadata.id-derived domain parts
  • Regenerate the tool catalog artifact (deterministically)
  • Run the sync generator that updates barrel exports/registries
  • Restart any long-running processes serving tool lists if the UI can be stale until restart
• Validation levels: declare what “done” means (contract-only vs runtime smoke vs staging).
  • Contract-only is valid for MVPs, but it must be labeled to avoid “implemented but not runnable” confusion.

External Research Triggers

Ask these questions. If any answer is "yes", research before proceeding:

Question	If Yes...
Does this plan use patterns/frameworks not in the codebase?	Research latest documentation and examples
Does this integrate with external services I haven't used?	Get API docs, authentication flows, rate limits
Is this the first implementation of a new capability type?	Find reference implementations
Are there security/compliance requirements I'm uncertain about?	Research standards and best practices
Has the underlying technology changed since I last used it?	Check for breaking changes, deprecations

External Research Template

### External Knowledge Needed

| Topic               | Question                     | Research Source         | Findings          |
| ------------------- | ---------------------------- | ----------------------- | ----------------- |
| [Framework/Pattern] | How to implement X?          | Official docs, examples | [Key patterns]    |
| [Integration]       | How does Y API work?         | OpenAPI spec, tutorials | [Auth, endpoints] |
| [Best Practice]     | What's the current standard? | Web search, community   | [Recommendations] |

### Confidence After Research

| Area                    | Before          | After           | Notes |
| ----------------------- | --------------- | --------------- | ----- |
| Implementation approach | Low/Medium/High | Low/Medium/High |       |
| Integration feasibility | Low/Medium/High | Low/Medium/High |       |
| Security posture        | Low/Medium/High | Low/Medium/High |       |

When to Skip Phase 0

• All patterns are already in the codebase
• All integrations have existing capabilities
• High confidence in implementation approach

---

Phase 1: Multi-Persona Evaluation

Evaluate the plan from 5 perspectives. For each persona, score alignment (1-10) and identify gaps.

Scoring Rubric

Use this rubric for consistent scoring across personas:

Score	Meaning	Criteria
9-10	Excellent	All needs addressed, patterns documented, no gaps identified
7-8	Good	Most needs addressed, minor gaps can be resolved during execution
5-6	Adequate	Core needs addressed, but significant gaps require pre-work
3-4	Weak	Major concerns, substantial pre-work or plan revision needed
1-2	Poor	Fundamental misalignment, plan needs significant rethinking

Readiness Thresholds

Average Score	Readiness	Action
≥7.0	Ready	Proceed with execution
5.0-6.9	Needs Pre-Work	Complete identified pre-work, then proceed
<5.0	Needs Rethink	Revise plan before investing in pre-work

Required Personas

1. Agent (AI Assistant)

Key Concerns: Tool discoverability, determinism, context propagation, HITL clarity

Evaluation Criteria	Questions to Ask
Tool Discovery	Are all new capabilities registered in MCP? Will agents find them?
Determinism	Does workflow code follow WCS-001 determinism rules?
Context Propagation	Is GoldenContext properly extended and propagated?
HITL Clarity	Are approval points clearly defined with signals?

Generative Prompts for Agent Evaluation:

Imagine you are an AI agent trying to execute this plan's outcomes.
Walk through each phase and ask:

1. "When a user asks me to [action from this plan], what tool do I invoke?"
   - If unclear → Gap: MCP registration or tool naming

2. "What information do I need to pass to each capability?"
   - If unclear → Gap: Schema documentation or aiHints

3. "If something fails mid-workflow, can I retry safely?"
   - If no → Gap: Idempotency or determinism violation

4. "When do I need human approval before proceeding?"
   - If unclear → Gap: HITL signal definitions

5. "How do I know if the operation succeeded?"
   - If unclear → Gap: Output schema or status indicators

6. "If I surface an approval request to a human, what context do I include so it is reviewable?"
   - If unclear → Gap: execution/request contract missing incident/workflow context (eventId/incidentId, service tags, contextType)

7. "Will the tool actually appear where agents look for it?"
   - If unclear → Gap: deterministic tool-catalog regeneration + registry/barrel export sync step missing from plan

8. "If the tool list in the UI is stale, do we have a clear restart/refresh step?"
   - If unclear → Gap: runbook/action missing; add explicit “restart/refresh tool catalog” step for local/dev workflows

2. Developer (Platform Contributor)

Key Concerns: Dogfooding, clear patterns, fast feedback, minimal boilerplate

Evaluation Criteria	Questions to Ask
Dogfooding	Have we used the platform to build the platform?
Clear Patterns	Do skills exist for the patterns we're using?
Fast Feedback	Are tests comprehensive and fast?
Minimal Boilerplate	Do generators exist for repetitive scaffolding?

Generative Prompts for Developer Evaluation:

Imagine you are a new developer implementing a similar feature next month.
Walk through the plan and ask:

1. "What patterns will I use that aren't documented?"
   - If any → Gap: Skill creation needed

2. "What files will I create more than twice?"
   - If any → Gap: Generator needed

3. "How will I know if my code follows the standards?"
   - If unclear → Gap: Linting rules or TDD guidance

4. "Where will I get stuck waiting for information?"
   - Each wait → Gap: Documentation or spec needed

5. "What would I copy-paste from this implementation?"
   - Each item → Gap: Should be a pattern, not copy-paste

6. "Are tests deterministic across machines and CI?"
   - If unclear → Gap: avoid reliance on global/user-local directories in tests; add explicit overrides/fixtures

7. "If there are two execution surfaces (in-process vs container/runtime), do we keep them aligned?"
   - If unclear → Gap: missing smoke/contract tests proving both surfaces match on key invariants

8. "Are there any `${...}`-shaped template hazards in generated scripts or embedded strings?"
   - If yes → Gap: add explicit escaping conventions and tests for template/script output

3. End User (Platform Operator)

Key Concerns: Usability, response speed, clear workflows, minimal training

Evaluation Criteria	Questions to Ask
Usability	Is the UI intuitive? Are user flows documented?
Response Speed	Are SLOs defined? Is performance measured?
Clear Workflows	Are procedures documented? Do runbooks exist?
Minimal Training	Is there an onboarding guide?

Generative Prompts for End User Evaluation:

Imagine you are an SRE at 3am trying to use this system.
Walk through the plan and ask:

1. "What will I click/type to accomplish the main task?"
   - If unclear → Gap: UX design or user flow needed

2. "How long will I wait for results?"
   - If unknown → Gap: SLO definition needed

3. "What do I do if it doesn't work?"
   - If unclear → Gap: Error handling o

---

*Content truncated.*