SKILL_ROUTING_GUIDE.md

Agile V Skills: Routing Guide

Repository version: 3.9.x Catalog: 49 skills currently present. [Preview] means the skill's current YAML frontmatter contains metadata.status: draft; presence on main does not make that contract stable.

Load agile-v-core first, then only the stage and risk-relevant skills. See Installation Profiles and the Golden Journey.

Complete Catalog

Intent	Skill	Location	Status
Foundation, directives, context and lifecycle rules	agile-v-core	agile-v-core/	Current
Multi-agent waves and handoffs	agile-v-pipeline	agile-v-pipeline/	Current
Cycle 2+, change requests, archival	agile-v-lifecycle	agile-v-lifecycle/	Current
Risk, CAPA, approvals, security, revalidation	agile-v-compliance	agile-v-compliance/	Current
Runtime control matrix	agile-v-control-matrix	agile-v-control-matrix/	Current
Risk-based DQ/IQ/OQ/PQ qualification, intended-use validation, stage release, requalification	agile-v-gxp-qualification	agile-v-gxp-qualification/	[Draft]
Bainbridge-aware Human Oversight Case for L2+ Human Gates	agile-v-human-oversight	agile-v-human-oversight/	[Draft]
Bind controlled SOPs to Agile-V controls/gates (SOP conformance)	agile-v-sop-adapter	agile-v-sop-adapter/	[Draft]
Interface, test, data-type, and time-allocation checks	agile-v-quality-gates	agile-v-quality-gates/	Current
Coding anti-pattern prevention	agile-v-behavioral	agile-v-behavioral/	Current
AI run provenance and AI/ML-BOM	agile-v-aibom	agile-v-aibom/	[Preview]
Architecture Decision Records (immutable, supersession)	agile-v-adr	agile-v-adr/	[Draft]
User research to candidate requirements	discovery-analyst	discovery-analyst/	Current
Security/privacy threat analysis	threat-modeler	threat-modeler/	Current
UX, accessibility, interaction constraints	ux-spec-author	ux-spec-author/	Current
Formal requirements and baselines	requirement-architect	requirement-architect/	Current
Independent ambiguity/constraint findings	logic-gatekeeper	logic-gatekeeper/	Current
Language-agnostic implementation	build-agent	build-agent/	Current
Python implementation	build-agent-python	domains/build-agent-python/	Current
JavaScript/TypeScript/Web implementation	build-agent-js	domains/build-agent-js/	Current
NestJS implementation	build-agent-nestjs	domains/build-agent-nestjs/	Current
Dart/Flutter implementation	build-agent-dart	domains/build-agent-dart/	Current
Embedded C/C++ and firmware	build-agent-embedded	domains/build-agent-embedded/	Current
Hardware schematics, netlists, HDL	schematic-generator	schematic-generator/	Current
Independent test design from baseline	test-designer	test-designer/	Current
Independent execution and verification	red-team-verifier	red-team-verifier/	Current
Representative intended-use validation	validation-agent	validation-agent/	Current
Hazard analysis and safety assurance	safety-engineer	safety-engineer/	Current
Decision logs, traceability, audit evidence	compliance-auditor	compliance-auditor/	Current
Standards-based repository documentation	documentation-agent	documentation-agent/	Current
Backlog, sprint, and Product Owner work	agile-v-product-owner	agile-v-product-owner/	Current
Rollout, rollback, and deployment plan	release-manager	release-manager/	Current
Metrics, dashboards, alerts, SLOs	observability-planner	observability-planner/	Current
Existing-system Gate 0 overview	system-understanding-agent	skills/system-understanding-agent/	Current
Pre-change impact map	impact-analysis-agent	skills/impact-analysis-agent/	Current
Impact-based regression selection	regression-selection-agent	skills/regression-selection-agent/	Current
Requirement/component/test graph links	graph-traceability-agent	skills/graph-traceability-agent/	Current
Predicted-versus-actual diff evidence	diff-evidence-agent	skills/diff-evidence-agent/	Current
Business vision, model, portfolio	venture-strategist	venture-strategist/	[Preview]
R&D pipeline, radar, prototypes, IP	rd-innovator	rd-innovator/	[Preview]
Go-to-market, launch, growth	gtm-executor	gtm-executor/	[Preview]
Finance, OKRs, resources, vendors	business-operations	business-operations/	[Preview]
Shared C-Suite governance primitives	c-suite-foundation	c-suite-foundation/	[Preview]
CEO orchestration and strategic alignment	chief-exec	chief-exec/	[Preview]
CTO architecture governance	chief-tech	chief-tech/	[Preview]
CFO financial governance	chief-finance	chief-finance/	[Preview]
CHRO people governance	chief-people	chief-people/	[Preview]
COO operational governance	chief-ops	chief-ops/	[Preview]
Periodic executive briefing aggregation	c-suite-update	c-suite-update/	[Preview]

Key Routing Distinctions

User asks for	Route	Boundary
“Check the requirements”	logic-gatekeeper	Records independent findings; never rewrites the draft or baseline
“Build this”	build-agent + one domain build skill	Requires Gate 1-approved, frozen baseline
“Design the tests”	test-designer	Reads the baseline, not implementation
“Verify it works as specified”	red-team-verifier	Produces .agile-v/VERIFICATION_SUMMARY.md and VER-XXXX evidence
“Is this human approval real oversight, or just a signature?”	agile-v-human-oversight	Requires a blind precommit, claim-specific independence, and recovery evidence — a second agent is not independent assurance
“Validate it with users/in operations”	validation-agent	Produces intended-use validation plan/protocol/report after verification
“Run Eval Gate / prepare Gate 2”	red-team-verifier + compliance-auditor	EVAL_RESULTS.md status must be PASS or authorized WAIVED; link it from VERIFICATION_SUMMARY.md
“Assess hazards or unacceptable harm”	safety-engineer	Tailors safety methods and assurance; does not replace verification or validation
“Can the agent/tool execute?”	agile-v-control-matrix + agile-v-compliance	Check data, tool, model, rights, approval, rollback, and evidence controls
“Understand this existing repo/change impact”	system-understanding-agent then impact/regression skills	Complete Gate 0 before requirements and build
“Plan a release”	release-manager	Starts after Gate 2 approval; no autonomous production release
“Set strategy/govern the business”	relevant preview business/C-Suite skill	Preview contracts require local approval and baselining

Canonical Workflows

New or Changed Feature

agile-v-core -> risk classification -> requirement-architect (persist draft) -> logic-gatekeeper (independent findings) -> architect revision -> Human Gate 1 -> frozen baseline -> build-agent + test-designer (independent) -> red-team-verifier -> Eval Gate -> validation-agent when intended-use validation is required -> AI manifest/evidence bundle -> Human Gate 2 -> release-manager -> observability-planner

Follow the evidence and stop conditions in the Golden Journey.

Existing Repository

1. system-understanding-agent creates the Gate 0 system overview.
2. impact-analysis-agent predicts affected components, interfaces, risks, and tests.
3. regression-selection-agent selects existing regression coverage and flags gaps.
4. Continue the canonical feature flow through a persisted draft and Gate 1 baseline.
5. After implementation, graph-traceability-agent links requirements, components, changes, and tests.
6. diff-evidence-agent compares predicted and actual impact before independent verification and Gate 2.

Regulated or Safety-Relevant Work

Use the regulated install profile. Classify L0-L4; add threat-modeler, safety-engineer, and validation-agent according to scope; maintain control, trace, approval, verification, applicable validation, release, and AI influence evidence. Load agile-v-human-oversight at Human Gates for L2+ work so approval is oversight-effectiveness evidence, not only authority evidence. Skills and templates support an assurance process but do not establish certification or regulatory conformity.

Business Preview

Load c-suite-foundation before a relevant chief-* skill. C-Suite skills govern and delegate; functional skills execute strategy, R&D, GTM, and operations. Use c-suite-update only to aggregate already controlled domain evidence. Do not treat preview outputs as approved decisions without accountable human review.

Loading Rules

1. Load agile-v-core first.
2. Use the smallest applicable installation profile.
3. Preserve independence with separate fresh contexts; if one context is unavoidable, design tests before implementation.
4. Pass durable file references, not chat-only handoffs.
5. Requirement changes after Gate 1 require a change request, independent review, approval, and a new baseline.
6. Keep verification and intended-use validation evidence separate.
7. A skill's frontmatter status controls preview labeling; branch location does not override it.