Nova Route

Shared Execution Policy

This file is the supporting behavioral contract for /nova-plugin:route and the deprecated /nova-plugin:nova-route compatibility entrypoint. Prefer the direct command; the compatibility name remains only for the current major-version migration window.

• Resolve natural-language and explicit KEY=value inputs using ../_shared/parameter-resolution.md; explicit non-conflicting values take precedence.
• Apply ../_shared/safety-preflight.md before side effects. Never infer approval, destructive scope, credentials, or output destinations.
• Follow ../_shared/output-contracts.md and ../_shared/artifact-policy.md; report completed, skipped, and blocked validation truthfully.
• Respect the frontmatter tool boundary. Missing inputs, unavailable dependencies, overlapping user changes, or repository-policy conflicts are blockers rather than permission to broaden scope.

Execution

1. Parse $ARGUMENTS against the workflow-specific inputs below.
2. Read only the context required for the requested scope.
3. Apply the workflow contract and its strict output format.
4. Stop before unauthorized side effects; otherwise validate in proportion to risk and report residual risk.

Workflow Contract

Generated from workflow-specs/behaviors.v2.json. This block is authoritative. Run node scripts/generate-behavior-surfaces.mjs --write after changing the IR; if explanatory text below conflicts, fail closed.

Generated Behavior Index

• Purpose: Choose the shortest safe next workflow route before execution starts.
• Canonical inputs: REQUEST(required aliases=INPUT,INTENT); DEPTH(optional aliases=MODE default="normal" exact="brief","normal")
• Resolved variant authority: {} normalized={} -> runtime/contracts/route.json. Declared selector defaults are applied before matching. An exact normalized override wins; a non-exact combination that triggers an alias specialization stops as conflicting, and only a valid combination that triggers no specialization uses the canonical fallback. The complete resolved runtime contract is authoritative and no field falls back to canonical prose.
• Claude static-entrypoint gate: Native command and Skill frontmatter are static. A matching command wrapper may continue after it has verified that its invoked command id equals resolvedWorkflowId; this canonical Skill must not re-resolve or reject that validated wrapper. Only when this canonical Skill is itself the Claude native invoked entrypoint and no validated wrapper gate exists must resolvedWorkflowId equal route. Otherwise STOP before tools or side effects and invoke the exact direct command /nova-plugin:<resolved commandEntrypoint.directCommandId>; never execute the specialized contract under unmatched canonical frontmatter. Generic and Codex adapters may execute the resolved contract directly under adapter enforcement.
• Decision entries: 19; canonical routes and variants: implement-plan {"EXECUTION_PROFILE":"codex-review-fix"}, review {"REVIEW_PROFILE":"codex-review-only"}, review {"REVIEW_PROFILE":"codex-verify-only"}, explore {"DEPTH":"deep"}, explore {"PERSPECTIVE":"reviewer"}, explore {}, produce-plan {"PLAN_PROFILE":"java-backend"}, review {"REVIEW_PROFILE":"plan"}, produce-plan {"PLAN_PROFILE":"lite"}, produce-plan {}, review {"LEVEL":"strict"}, review {"LEVEL":"lite"}, review {"LEVEL":"standard","MODE":"findings-only"}, review {}, implement-plan {}, implement-plan {"EXECUTION_PROFILE":"lite"}, implement-plan {"EXECUTION_PROFILE":"standard"}, finalize-work {"DEPTH":"lite"}, finalize-work {}.
• Workflow steps: resolve-intent → classify → select → verify-surface → emit
• Output: mode=chat; order=Canonical skill → Command entrypoint → Variant parameters → Core agent → Capability packs → Required inputs → Validation expectations → Fallback path; severity=none.
• Deviation/failure: mode=forbid; failure order=status → ambiguous intent → required choice → safe fallback.
• Full IR: runtime/contracts/route.json#behaviorContract embeds the complete decision table, invariants, stops, field definitions, validation, and failure contract from the same source. Detailed guidance below may not override it.

Purpose

Choose the next workflow step before work starts. It improves routing for agents that do not invoke Claude Code slash commands.

Inputs

Route Selection

Use the generated decision entries above as the complete route table. Select one canonical target, validate its variant parameters, and report the exact matching direct command. Specialized and compatibility commands are direct entrypoints, not extra route identities. Claude redirects when static frontmatter differs; generic and Codex adapters may execute the resolved contract under adapter enforcement.

Output Format

## Recommended Route

- Canonical skill:
- Command entrypoint:
- Variant parameters:
- Core agent:
- Capability packs:
- Required inputs:
- Validation expectations:
- Fallback path:

Required inputs names the selected downstream workflow's canonical inputs, not the route command's own REQUEST. Use the exact UPPER_SNAKE_CASE names from the generated behavior contract and runtime contract. Always list the complete ordered required-input set even when values are already present, inferable, or resolved; never list only unresolved inputs. Do not translate the names into prose or substitute aliases.

For DEPTH=normal, include a one-sentence rationale inside the appropriate fixed field. For DEPTH=brief, keep every field concise. Do not add content outside the heading and eight fixed bullets.

Always output exactly one immediate next step. If the request spans later stages, describe those stages only as prose inside Validation expectations or Fallback path; do not add another canonical skill, command entrypoint, or variant-parameter identity.

Routing Rules

• Emit exactly one immediate next canonical route and matching command entrypoint.
• Automatic routing selects only one of the six canonical workflow targets; compatibility entrypoints are never the selected route identity.
• For read-only or findings-only intent, select canonical review with {"LEVEL":"standard","MODE":"findings-only"} and report /nova-plugin:review-only as the exact Claude command entrypoint.
• Use only existing capability packs: java, security, dependency, docs, release, marketplace, frontend, and mcp.
• Route source-grounding and official-documentation work through docs, mcp, or the relevant domain pack.
• Route doubt-driven high-risk review through security, dependency, or the relevant domain pack, with reviewer ownership.
• Route deprecation and migration planning through release, with architect and publisher ownership as needed.
• Do not invent commands, skills, packs, agents, validation claims, or implementation steps.

Safety

• Do not modify project files.
• Do not create route artifacts unless a future command explicitly adds a write-capable route variant.
• Do not run implementation, test, Git, install, network, or external review commands.
• Do not claim a validation command passed; only name what should be run by the selected downstream workflow.

Verification

• REQUEST was resolved or the missing input was named.
• The selected route is one of the six product-declared canonical targets; the exact matching command entrypoint is present, and variant parameters validate against the generated selector schema and resolution rule.
• The core agent is one of orchestrator, architect, builder, reviewer, verifier, or publisher.
• Capability packs are selected only from the existing pack set.
• Required inputs use the selected workflows' exact canonical UPPER_SNAKE_CASE names and validation expectations are explicit.
• The output is read-only and does not include implementation work.