Qperspective

Qperspective — Multi-Perspective Analysis

Role

Analyzes a single problem from multiple independent viewpoints simultaneously. Each perspective runs as a separate sub-agent, ensuring genuinely independent analysis without contamination between viewpoints.

CLI Interface

/Qperspective <topic>                         # Auto-select relevant perspectives
/Qperspective <topic> --views "dev,pm,user"   # Specify perspectives manually
/Qperspective <topic> --codex                 # Add Codex as an independent perspective
/Qperspective --help                          # Show usage

Available Perspectives

Perspective	Asks	Best for
developer	"Is this implementable? What's the technical cost?"	Architecture, tech debt decisions
user	"Does this actually solve my problem?"	Feature prioritization, UX
pm	"Does this fit the roadmap? What's the ROI?"	Scope, resource allocation
security	"What can go wrong? What's the attack surface?"	Auth, data handling, infra
business	"Does this make money? What's the market risk?"	Strategy, pricing, GTM
ops	"Can we run this in production? What breaks at scale?"	Deployment, monitoring
newcomer	"I don't understand any of this. Explain it simply."	Complexity check, onboarding
critic	"This will fail because..."	Risk assessment, blind spots
codex	Independent analysis from a different model	Cross-model validation

Execution Procedure

Step 1: Parse input

• Extract the topic/problem
• If --views specified, use those perspectives
• If not, auto-select 3-4 most relevant perspectives based on the topic
• If --codex flag, add Codex as an additional perspective

Step 2: Confirm perspectives

Topic: "Should we migrate from REST to GraphQL?"

Selected perspectives:
  1. developer  — technical feasibility and migration cost
  2. user       — API consumer experience
  3. ops        — production complexity and monitoring
  4. codex      — independent cross-model analysis

Proceed? [Y/n]  (or specify: --views "dev,security,pm")

Step 3: Run perspectives in parallel

Spawn one sub-agent per perspective, all running simultaneously:

Each agent receives:

• The topic/problem statement
• Their assigned perspective and guiding question
• Instruction to analyze ONLY from their viewpoint
• No access to other perspectives' outputs

For Codex perspective:

• Route via codex-plugin-cc using resolveEngine
• Falls back to an additional Claude sub-agent if unavailable

Each agent outputs:

• Assessment (3-5 key points from their viewpoint)
• Risks they see
• Opportunities they see
• One recommendation

Step 4: Display results

Multi-Perspective Analysis
──────────────────────────

Topic: "Should we migrate from REST to GraphQL?"

┌─ Developer ──────────────────────────────────┐
│ Assessment:                                   │
│   - Migration cost: ~3 sprints for core APIs  │
│   - N+1 query risk without DataLoader         │
│   - Type safety improves with codegen         │
│ Risk: Schema design mistakes are hard to fix  │
│ Opportunity: Eliminate over-fetching on mobile │
│ Recommendation: Start with BFF, not full swap │
└───────────────────────────────────────────────┘

┌─ User ───────────────────────────────────────┐
│ Assessment:                                   │
│   - Current REST API requires 3 calls for ... │
│   ...                                         │
└───────────────────────────────────────────────┘

┌─ Ops ────────────────────────────────────────┐
│   ...                                         │
└───────────────────────────────────────────────┘

┌─ Codex (independent) ────────────────────────┐
│   ...                                         │
└───────────────────────────────────────────────┘

Step 5: Cross-perspective synthesis

After displaying all perspectives:

Synthesis
─────────

Agreement across perspectives:
  - ...

Conflicts between perspectives:
  - Developer vs Ops: ...
  - User vs Business: ...

Blind spots (not covered by selected perspectives):
  - ...

Suggested next step:
  - ...

Step 6: Follow-up options

• "Want to dive deeper into one perspective?"
• "Should I run /Qdebate on the main conflict?"
• "Draft a decision document based on this analysis?"

Will

• Run 3-6 perspectives in parallel as independent sub-agents
• Support Codex as an independent cross-model perspective
• Produce a cross-perspective synthesis highlighting agreements and conflicts
• Suggest follow-up actions

Will Not

• Run more than 6 perspectives (noise exceeds signal)
• Mix perspectives within a single agent (defeats independence)
• Make the decision for the user
• Replace Qdebate for deep back-and-forth on a specific tradeoff