Sacco research to plan
Skill adrien-sacco/sacco-skills/skills/sacco-research-to-plan
Research-to-plan pipeline: dispatches multi-source research, produces structured requirements with reference IDs, derives phased architecture plan. Use when exploring a new domain, designing a new system, evaluating technology options, or when the user says "research", "figure out the best way", "look at best practices", "state of the art". Integrates with a deep-research tool for source dispatch (if available) and sacco-phased-execution for plan handoff.From its SKILL.md
npx -y skills add adrien-sacco/sacco-skills --skill sacco-research-to-planAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.9 KB, as published. Nobody here has run it
Research-to-Plan Pipeline
Transforms a research question into an actionable, phased implementation plan with full traceability from research findings to requirements to architecture to phases.
When to Use
- Starting a new project or major feature
- Technology selection (choosing frameworks, crates, services)
- Exploring a domain (e.g. a new data platform, ML architectures, etc.)
- User asks to "research and plan" something
- User says "figure out the best way to..."
Pipeline Stages
Stage 1: Research Dispatch
Use a deep-research tool/skill to launch multi-source research, if available. Otherwise dispatch 3-5 parallel research agents, each focused on a different facet:
Agent 1: State-of-the-art in <domain> (academic + industry)
Agent 2: Existing tools/libraries/services that solve this
Agent 3: Architecture patterns and best practices
Agent 4: Case studies and production deployments
Agent 5: Cost/performance/trade-off analysis
Output: docs/research/<topic>-research.md with source citations
Stage 2: Requirements Extraction
From research findings, produce a requirements table:
# Requirements: <feature/system>
| Ref | Category | Requirement | Priority | Source | Notes |
|-----|----------|------------|----------|--------|-------|
| R-001 | Core | Must support graph traversal with 1M+ nodes | Must | research-agent-3 | Comparable systems handle 10M+ |
| R-002 | Core | Real-time incremental updates | Must | research-agent-1 | Streaming-update pattern |
| R-003 | UX | Canvas-based node interaction | Should | research-agent-4 | Node-link diagram UX |
| R-004 | Ops | Horizontal scalability | Could | research-agent-5 | Only if >100 concurrent users |
Categories: Core, UX, Data, Security, Ops, Integration Priorities: Must (MVP), Should (v1.1), Could (backlog), Won't (explicit exclusion)
Output: docs/plans/<topic>-requirements.md
Stage 3: Architecture Derivation
From requirements, derive the architecture:
- Component map: What components are needed to satisfy each requirement
- Technology selection: For each component, which library/framework/service
- Must justify: why this over alternatives?
- Must cite: what research finding supports this choice?
- Data flow: How data moves between components
- API contracts: Key interfaces between components
- Non-functional requirements: Performance targets, scaling strategy
Output: docs/plans/<topic>-architecture.md
Stage 4: Phased Plan
Break the architecture into phases for sacco-phased-execution:
# Implementation Plan: <feature/system>
## Phase 0: Foundation
- **Deliverables:** Project scaffold, CI, basic types
- **Requirements covered:** (none — infrastructure only)
- **Estimated tests:** 20-30
- **Dependencies:** none
## Phase 1: <core-capability>
- **Deliverables:** <specific components>
- **Requirements covered:** R-001, R-002
- **Estimated tests:** 50-80
- **Dependencies:** Phase 0
## Phase 2: <next-capability>
...
Rules for phasing:
- Each phase must be independently testable and committable
- Must requirements come first (Phases 0-N), Should requirements after, Could last
- Each phase lists which requirements it satisfies (traceability)
- Phase 0 is always foundation/scaffold
Output: docs/plans/<topic>-plan.md
Stage 5: Handoff
Present the complete pipeline output to the user:
- Research summary (key findings, not the full report)
- Requirements table
- Architecture overview
- Phased plan
Wait for user approval before execution. The plan feeds directly into sacco-phased-execution.
Rules
- Research MUST happen before requirements. No inventing requirements from assumptions.
- Every requirement must trace to a research finding or explicit user input.
- Every architecture decision must trace to a requirement.
- Every phase must trace to requirements it covers.
- If research reveals the problem is already solved (existing tool/library), recommend that instead of building.
- The plan must include explicit "Won't" items to prevent scope creep.
Integration
- A deep-research tool (if available): Source dispatch in Stage 1
- sacco-phased-execution: Plan handoff in Stage 5
- sacco-codebase-audit: Can audit the plan before execution begins
- sacco-architecture-evaluation: Can be used in Stage 3 for architecture decisions
- Research-before-building: This skill is how you avoid solving already-solved problems