// Constitutional governance for PikaKit: non-negotiable laws for correctness, trust, and durability. Use when reviewing breaking changes, critical data operations, or governance violations. NOT for general code review (use code-review) or security scanning (use security-scanner).
AI-optimized browser automation with context-efficient @ref element handles. Use when automating browser interactions for AI agents with minimal context consumption. NOT for manual E2E testing (use e2e-automation) or screenshots (use chrome-devtools).
Prompt engineering for AI image generation: Midjourney, DALL-E, Stable Diffusion, Flux. Use when writing image prompts, configuring generation parameters, or optimizing AI art output. NOT for UI design decisions (use design-system) or media file processing (use media-processing).
API design patterns for REST, GraphQL, and tRPC services. Use when designing endpoints, versioning APIs, or choosing API architecture. NOT for code implementation (use nodejs-pro/python-pro) or database schema (use data-modeler).
Authentication and authorization: OAuth2, JWT, RBAC/ABAC, MFA, Passkeys, session management. Use when implementing login, permissions, tokens, or security flows. NOT for vulnerability scanning (use security-scanner) or API design (use api-architect).
Puppeteer CLI scripts for browser screenshots, debugging, and Core Web Vitals measurement. Use when capturing screenshots, measuring performance, or debugging browser issues. NOT for E2E test suites (use e2e-automation) or AI browser automation (use agent-browser).
Production deployment workflows, CI/CD pipelines, rollback strategies, and release management. Use when deploying apps, configuring pipelines, or planning rollback procedures. NOT for git operations (use git-workflow) or Vercel-specific deploys (use vercel-deploy).
| name | code-constitution |
| description | Constitutional governance for PikaKit: non-negotiable laws for correctness, trust, and durability. Use when reviewing breaking changes, critical data operations, or governance violations. NOT for general code review (use code-review) or security scanning (use security-scanner). |
| metadata | {"author":"pikakit","version":"3.9.206","category":"governance","triggers":["architecture","critical data","breaking change","governance","doctrine"],"coordinates_with":["code-review","security-scanner","code-craft"],"success_metrics":["0 governance bypasses","100% doctrine compliance"]} |
Supreme authority. Zero-trust agents. Fail-closed enforcement. Doctrine-first decisions.
Status: OFFICIAL · LOCKED | Authority: SUPREME | Override: NONE
Required: None — Code Constitution is a governance framework with no external dependencies.
Doctrine library: 16 rule files in rules/ directory (required for evaluation).
| Scope | Activation | Default |
|---|---|---|
| System architecture or boundaries | Mandatory — always load | — |
| Data ownership or persistence | Mandatory | — |
| Critical business logic | Mandatory | — |
| Breaking changes or commercial risk | Mandatory | — |
| AI agent behavior or autonomy | Mandatory | — |
| Ambiguous intent | Mandatory — load by default | Refuse |
| Architecture review, contracts, security | See rules/engineering-spec.md | — |
| Owned by This Skill | NOT Owned |
|---|---|
| Agent operating mode (STRICT/PROPOSAL_ONLY/ZERO_TRUST) | Agent implementation |
Doctrine library (16 rules in rules/) | Rule execution in other skills |
| Enforcement behavior (Stop → Cite → Refuse) | Automated remediation |
| Authority hierarchy definition | Skill priority in GEMINI.md |
| Change Proposal review | Proposal tooling |
Pure decision skill: Produces governance decisions (approve/refuse/escalate). Zero side effects.
| Phase | Action | Output |
|---|---|---|
| Classify | Validate request type, extract scope, identify risk level | Validated input or error |
| Evaluate | Load applicable doctrines from rules/, check each against context | Violation list (may be empty) |
| Decide | 0 violations → approve; ≥1 blocking → refuse; ambiguous → escalate | Decision + conditions/violations |
| Emit | Return structured output with enforcement action | Complete output schema |
All phases synchronous. Fail-closed: any phase failure defaults to "refuse."
Code Constitution (SUPREME)
└── GEMINI.md (P0)
└── Agent .md files (P1)
└── Skill .md files (P2)
No skill, agent, or workflow may override a constitutional decision.
| Setting | Value | Meaning |
|---|---|---|
| POLICY_MODE | STRICT | All rules enforced without exception |
| AUTONOMY | PROPOSAL_ONLY | Agents propose; never self-approve |
| TRUST_LEVEL | ZERO | Agent declarations are inputs, not truth |
Agents MAY: Propose actions, explain trade-offs, identify risks, cite doctrines.
Agents MUST NOT: Approve without user consent, merge automatically, bypass enforcement.
When violation detected:
Fail-closed invariant: Partial compliance = enforcement failure. Ambiguity = refuse or escalate.
| Code | Recoverable | Default Action |
|---|---|---|
ERR_INVALID_REQUEST_TYPE | No | Refuse |
ERR_MISSING_CONTEXT | Yes | Refuse |
ERR_DOCTRINE_NOT_FOUND | No | Refuse |
ERR_AMBIGUOUS_SCOPE | Yes | Escalate |
ERR_TRUST_VIOLATION | No | Refuse + restrict |
ERR_INVALID_PROPOSAL | Yes | Refuse |
ERR_EVALUATION_FAILED | No | Refuse |
All failures default to refuse. No silent approvals. Zero internal retries.
1. Agent proposes action with context (scope, files, change type, risk level)
2. Code Constitution validates against applicable doctrines in rules/
3. Decision returned: approve | refuse | escalate
4. If refused: Stop, cite violation, do NOT proceed
| ❌ Don't | ✅ Do |
|---|---|
| Self-approve agent actions | Propose and wait for user consent |
| Ignore governance for "quick fix" | Load constitution; validate first |
| Override constitution from another skill | Respect authority hierarchy |
| Assume no violation = safe | Explicitly check applicable doctrines |
| Modify doctrines directly | Submit Change Proposal |
| Problem | Cause | Resolution |
|---|---|---|
| Every action refused | Scope set to "critical" for minor changes | Set accurate risk_level and scope |
| Doctrine not found error | Skill installation incomplete | Verify rules/doctrines/ contains all 8 domains |
| Escalation loop | Ambiguous scope keeps re-triggering | Provide explicit scope: architecture, data, security |
| Agent bypasses governance | Skill not loaded for the request type | Add trigger keywords to agent routing config |
| Folder | Content | When to Read |
|---|---|---|
| rules/ | 16 domain-specific governance rules | Doctrine evaluation |
| rules/engineering-spec.md | Full engineering spec: contracts, security, scalability | Architecture review |
| metadata/ | Routing & intent configuration | Skill routing |
| resources/ | Templates & reference materials | Building governance artifacts |
| examples/ | Governance usage examples | Learning patterns |
| knowledge/ | Accumulated governance knowledge | Context building |
| proposals/ | Change Proposals for doctrine updates | Modifying constitution |
| Item | Type | Purpose |
|---|---|---|
code-review | Skill | Code quality review |
security-scanner | Skill | Security vulnerability detection |
code-craft | Skill | Code standards enforcement |
skill-generator | Skill | Skill file validation |
⚡ PikaKit v3.9.206