Define requirements for an OpenClaw integration
Document a proposed integration's data needs, actions, credentials, and review boundaries.
Systematic analysis and reasoning workflows. Use when performing audits, investigation, requirements analysis, risk analysis, scenario/edge-case enumeration, or design analysis that demands evidence-based, comprehensive coverage.
---
name: analysis-design-e41a2616440f
description: Systematic analysis and reasoning workflows. Use when performing audits, investigation, requirements analysis, risk analysis, scenario/edge-case enumeration, or design analysis that demands evidence-based, comprehensive coverage.
allowed-tools: ["shell", "apply_patch", "read_file", "write_file", "update_plan"]
metadata:
short-description: Structured analysis workflow
---
# Analysis Design
**Read these reference files before proceeding:**
- `~/.config/agent/core/behavior.md` - Critical thinking and evidence requirements
- `~/.config/agent/core/methodology.md` - 5-phase problem-solving framework
- `~/.config/agent/core/efficiency.md` - Parallelization and throughput
- `~/.config/agent/core/task-management.md` - TODO discipline
- `~/.config/agent/core/principles.md` - Engineering principles
- `~/.config/agent/core/communication.md` - Communication standards
- `~/.config/agent/domain/architecture/decomposition.md` - Problem breakdown
- `~/.config/agent/domain/architecture/parallelization.md` - Parallel execution
- `~/.config/agent/domain/testing/unit.md` - Unit testing strategy
- `~/.config/agent/domain/testing/coverage.md` - Coverage expectations
## Instructions
Apply all guidance from the reference files listed above.
Use this workflow when the task requires rigorous, comprehensive analysis:
1. **Scope & Objective**
- Restate the objective and success criteria explicitly.
- Define in-scope and out-of-scope boundaries.
- Capture constraints and assumptions up front.
2. **Evidence Collection**
- Read relevant files and source-of-truth artifacts before reasoning.
- Record evidence with precise references (paths, symbols, tests).
- Mark each claim as Verified or Unverified based on evidence.
3. **Scenario/Edge-Case Matrix**
- Enumerate all viable scenarios/contexts (including non-goals).
- For each scenario, specify expected behavior and verification status.
- Tag missing coverage explicitly as TODO with a test plan.
4. **Risk & Unknowns**
- Identify unknowns and risks early; propose mitigations or decision points.
- Separate solvable risks from unsolvable ones; document residual risk.
5. **Recommendations & Plan**
- Provide prioritized recommendations (P0/P1/P2 or equivalent).
- Map each recommendation to evidence and a verification step.
- Update or produce a plan that reflects verified gaps and required policy decisions.
6. **Verification Strategy**
- Define how correctness, safety, and completeness will be validated.
- Prefer tests or reproducible checks; document gaps if verification is blocked.
**Deliverables (as applicable):**
- Evidence log (Verified vs Unverified)
- Scenario/edge-case matrix with TODOs
- Risk register with mitigation plan
- Updated implementation plan or decision record
## Arguments
Target: ${ARGUMENTS}
---
name: analysis-design-e41a2616440f
description: Systematic analysis and reasoning workflows. Use when performing audits, investigation, requirements analysis, risk analysis, scenario/edge-case enumeration, or design analysis that demands evidence-based, comprehensive coverage.
allowed-tools: ["shell", "apply_patch", "read_file", "write_file", "update_plan"]
metadata:
short-description: Structured analysis workflow
---
# Analysis Design
**Read these reference files before proceeding:**
- `~/.config/agent/core/behavior.md` - Critical thinking and evidence requirements
- `~/.config/agent/core/methodology.md` - 5-phase problem-solving framework
- `~/.config/agent/core/efficiency.md` - Parallelization and throughput
- `~/.config/agent/core/task-management.md` - TODO discipline
- `~/.config/agent/core/principles.md` - Engineering principles
- `~/.config/agent/core/communication.md` - Communication standards
- `~/.config/agent/domain/architecture/decomposition.md` - Problem breakdown
- `~/.config/agent/domain/architecture/parallelization.md` - Parallel execution
- `~/.config/agent/domain/testing/unit.md` - Unit testing strategy
- `~/.config/agent/domain/testing/coverage.md` - Coverage expectations
## Instructions
Apply all guidance from the reference files listed above.
Use this workflow when the task requires rigorous, comprehensive analysis:
1. **Scope & Objective**
- Restate the objective and success criteria explicitly.
- Define in-scope and out-of-scope boundaries.
- Capture constraints and assumptions up front.
2. **Evidence Collection**
- Read relevant files and source-of-truth artifacts before reasoning.
- Record evidence with precise references (paths, symbols, tests).
- Mark each claim as Verified or Unverified based on evidence.
3. **Scenario/Edge-Case Matrix**
- Enumerate all viable scenarios/contexts (including non-goals).
- For each scenario, specify expected behavior and verification status.
- Tag missing coverage explicitly as TODO with a test plan.
4. **Risk & Unknowns**
- Identify unknowns and risks early; propose mitigations or decision points.
- Separate solvable risks from unsolvable ones; document residual risk.
5. **Recommendations & Plan**
- Provide prioritized recommendations (P0/P1/P2 or equivalent).
- Map each recommendation to evidence and a verification step.
- Update or produce a plan that reflects verified gaps and required policy decisions.
6. **Verification Strategy**
- Define how correctness, safety, and completeness will be validated.
- Prefer tests or reproducible checks; document gaps if verification is blocked.
**Deliverables (as applicable):**
- Evidence log (Verified vs Unverified)
- Scenario/edge-case matrix with TODOs
- Risk register with mitigation plan
- Updated implementation plan or decision record
## Arguments
Target: ${ARGUMENTS}
Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.