Pair capability assess ai

/pair-capability-assess-ai — AI Development Assessment

Evaluate and recommend AI development tools: AI assistants, MCP integrations, AI-specific SDKs, and models. Follows the resolution cascade. Output-only: produces a proposal (rendered AI-section content + target) plus a report — writes no files. Persistence is delegated to /pair-capability-record-decision.

Arguments

Argument	Required	Description
$choice	No	Override: skip assessment, use this AI tool directly (e.g. claude-code, cursor)

Composed Skills

Output-only — composes no skill, writes no files; the caller persists via /pair-capability-record-decision (see Composition Interface).

Proposal Target

The rendered adoption content is destined for this section — the caller writes it via /pair-capability-record-decision:

• Target: adoption/tech/tech-stack.md — AI section only
• Ownership: AI section (shared file — /pair-capability-assess-stack owns core sections, /pair-capability-assess-testing owns testing section)

Algorithm

Step 1: Resolution Cascade

Read resolution cascade for the generic Path A/B/C mechanics (check → skip → act → verify).

• Path A delta: override argument is $choice. On confirm, proceed to Step 3.
• Path B delta: adoption check is adoption/tech/tech-stack.md — populated AI section. If a corresponding decision record is missing, report the gap (this skill still writes nothing; the caller persists a backfill via /pair-capability-record-decision).
• Path C delta: proceed to Step 2.

Step 2: Read Guidelines

1. Act: Read AI development guidelines:
   • AI Development README — maturity model, tool selection matrix, implementation strategies
   • AI Tools — tool comparison and configuration
   • MCP Integration — MCP protocol standards
2. Act: Read project context:
   • adoption/product/PRD.md — AI requirements, team size
   • adoption/tech/tech-stack.md — existing stack (AI tools must integrate)
3. Verify: Every file listed in 1-2 above has been read; any that's missing follows Graceful Degradation (ask the developer directly) instead of silently proceeding to Step 3.

Step 3: Evaluate Options

1. Act: Use the Tool Selection Decision Matrix from guidelines:

   • Evaluate tools on: Code Generation, Context Awareness, Architecture Review, Team Integration, Security/Privacy.
   • Consider: budget, team size, workflow integration, privacy requirements.

2. Act: Evaluate AI development areas:

   • Primary AI assistant: IDE-integrated tool (Cursor, Copilot, Claude Code, etc.)
   • Strategic AI: Architecture review, complex problem-solving tool
   • MCP integration: If applicable, MCP servers and protocol adoption
   • AI-specific libraries: SDKs, embedding tools, AI frameworks with versions

3. Act: Assess team's AI maturity level (from guidelines maturity model):

   • Level 1: Basic AI Assistance
   • Level 2: Integrated AI Workflows
   • Level 3: Strategic AI Architecture
   • Level 4: AI-Native Development

4. Act: Present recommendation:

   AI Development Recommendation:

   • Primary assistant: [tool] — [rationale]
   • Strategic AI: [tool] — [rationale]
   • MCP: [yes/no — details]
   • AI maturity target: Level [N]
   • Additional tools: [list with versions]

5. Verify: Developer approves.

Step 4: Render Adoption Proposal

1. Act: Render the AI section content — the ready-to-write body for tech-stack.md:
   • AI assistants with versions/tiers
   • MCP integrations if applicable
   • AI-specific SDKs and libraries with versions
   • Scope strictly to the AI section so the caller's write preserves all other sections (core, testing)
2. Verify: The rendered content and its target are ready to emit.

Step 5: Emit Proposal

1. Act: Emit the proposal to the caller:
   • content: the rendered AI-section body from Step 4
   • target: adoption/tech/tech-stack.md (AI section)
   • decision-metadata: $type: non-architectural, $topic: ai-development-tools, $summary: "[Primary tool] adopted as AI development assistant with [maturity level] target"
   • plus the human-facing report (see Output Format)
2. Verify: Proposal emitted — see record-decision invocation contract for the persistence contract (persistence is always the caller's responsibility, delegated to /pair-capability-record-decision).

Output Format

Follows the Decision Shape.

ASSESSMENT COMPLETE (output-only — no files written):
├── Domain:    AI Development
├── Path:      [Argument Override | Adoption Exists | Full Assessment]
├── Decision:  [primary tool + maturity level + additional tools]
├── Proposal:  [content rendered for tech-stack.md AI section]
├── Target:    adoption/tech/tech-stack.md (AI section)
├── Persist:   [caller composes /record-decision(content, target) → ADL]
└── Status:    [Proposal ready | Confirmed existing]

Composition Interface

See record-decision invocation contract for the generic tuple + Input/Output/Persistence shape.

When composed by /pair-process-bootstrap:

• Input: /pair-process-bootstrap invokes during Phase 2.
• Persistence: /pair-process-bootstrap composes /pair-capability-record-decision to write the AI section and record the ADL.

When invoked independently: the human (or agent) persists the proposal by composing /pair-capability-record-decision, then commits.

Edge Cases

• tech-stack.md exists but no AI section: Render content that adds the AI section; the caller's write preserves all other content.
• Project doesn't use AI tools: Render a minimal section noting "AI development tools not adopted — [reason]" for the caller to persist.
• Multiple AI assistants: Document primary and secondary with roles (e.g. primary for coding, secondary for architecture review).
• MCP adoption requires infrastructure changes: Recommend composing /pair-capability-assess-infrastructure for infra implications.

Graceful Degradation

See graceful degradation (guideline missing → ask developer for AI tool preferences directly) and record-decision contract (persistence unavailable → proposal stands as a report) for the standard scenarios. No additional cases.

Notes

• AI tool decisions are non-architectural → the caller records them as an ADL. Exception: if MCP integration fundamentally changes system architecture, the caller uses ADR.
• Section ownership: this skill renders content ONLY for the AI section of tech-stack.md. The single adoption writer is /pair-capability-record-decision.
• Version tracking: every AI tool includes version or tier (e.g. "Claude Code", "GPT-4o", "Cursor Pro").
• Educational content (AI development principles, best practices, WHY) stays in guidelines.