@aisdlc/core 4.8.0 → 4.9.1

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/dist/index.cjs CHANGED
@@ -14354,7 +14354,8 @@ var SKILL_EXPECTED_TOOLS = {
14354
14354
  "aisdlc-doctor-migrate": ["doctor", "migrate"],
14355
14355
  "aisdlc-release-compliance": ["verify", "kpi_release", "report_dashboard"],
14356
14356
  "aisdlc-security-sca-triage": ["verify_security", "verify_privacy", "verify_licenses"],
14357
- "aisdlc-bdd-gherkin": ["sdd_check_fix", "verify_testing"]
14357
+ "aisdlc-bdd-gherkin": ["sdd_check_fix", "verify_testing"],
14358
+ "aisdlc-spot-audit": ["inspect_element"]
14358
14359
  };
14359
14360
  function readExecutedMcpTools(auditLogFile) {
14360
14361
  const executed = /* @__PURE__ */ new Set();
@@ -19354,6 +19355,29 @@ When consolidating multiple increments in a single release or PR, apply **only o
19354
19355
  1. Extract: \`npx aisdlc gherkin extract --all\`
19355
19356
  2. Scaffold RED steps: \`npx aisdlc gherkin scaffold <feature-file> --lang <lang> --fallback\`
19356
19357
  3. Verify sync: \`npx aisdlc verify gherkin\`
19358
+ `
19359
+ }
19360
+ ],
19361
+ "aisdlc-spot-audit": [
19362
+ {
19363
+ filename: "polymorphic-audit-matrix.md",
19364
+ content: `# Polymorphic Element Audit & Adversarial Matrix
19365
+
19366
+ ## Layer-to-Verifier & Owner Role Routing
19367
+ | Target Element / Layer | Deterministic Pre-Checks | \`agent-auditor\` Focus (Quality / YAGNI) | \`agent-adversary\` Focus (Red Team / PoC) | Return Handoff Owner Role |
19368
+ | :--- | :--- | :--- | :--- | :--- |
19369
+ | **Product** (\`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`, \`JRN-*\`, \`ARCH-INTRO-*\`) | \` inspect_element\`, \`verify_schemas\`, \`verify_duplicates\`, \`verify_pdac\` | Ambiguity, missing Gherkin \`@nominal\`/\`@boundary\`/\`@invalid\`, anti-library-only MVP & CRUD completeness | Contradictory rules, unreachable states, logic traps, unhandled exception flows | \`agent-product-analyst\` (or \`agent-expert-user\`) |
19370
+ | **Cybersecurity** (\`ACT-THREAT-*\`, \`ABUSE-*\`, \`SEC-REQ-*\`) | \`inspect_element\`, \`verify_schemas\`, \`verify_pdac\` | STRIDE/OWASP ASVS completeness, Zero Trust enclave binding (\`SEC-ENC-*\`) | Bypass vectors, unmitigated abuse paths, missing \`@security @mitigation\` negative BDD | \`agent-threat-modeler\` |
19371
+ | **Privacy** (\`DATA-ASSET-*\`, \`PRIV-ROPA-*\`, \`PRIV-REQ-*\`) | \`inspect_element\`, \`verify_schemas\`, \`verify_privacy\` | LINDDUN / GDPR Art. 30 completeness, retention & erasure strategy, DPIA gate | PII leakage vectors in logs/LLM prompts, re-identification / linkability attacks | \`agent-privacy-officer\` |
19372
+ | **Architecture** (\`CMP-*\`, \`ADR-*\`, \`ARCH-DEP-*\`, \`SEC-ENC-*\`) | \`inspect_element\`, \`verify_schemas\`, \`verify_traceability\` | Hexagonal Ports & Adapters, YAGNI, Contract-First DTOs, persistent \`CMP-OBSERVABILITY\` | SPOFs, race conditions, cascading failures, resource exhaustion, enclave boundary bypass | \`agent-system-architect\` |
19373
+ | **Verification Suites** (\`tests/**\`, \`*.feature\`) | \`inspect_element\`, \`verify_testing\`, \`npx aisdlc verify gherkin\` | Deterministic Gherkin sync, Zero-Mock integration, per-interactive-element Happy/Sad/Boundary coverage | Conceptual mutation testing (proving how broken/empty production code passes tests green) | \`agent-qa-engineer\` |
19374
+ | **Source Code** (\`src/**\`, \`packages/**\`) | \`inspect_element\`, \`verify_quality\`, \`verify_security\`, \`verify_privacy\`, \`verify_licenses\` | CC <= 10, Cognitive <= 15, MI >= 50, Lines <= 40, SOLID/DRY, 1:1 unit & UI tests, 4 UI states, zero fake-wiring | Boundary/injection PoC breaks (SQLi, XSS, path traversal, prompt injection), race triggers, CVSS v3.1 scoring | \`agent-developer\` |
19375
+ | **CI/CD & Infra** (\`.github/workflows/**\`, \`Dockerfile*\`, \`scripts/**\`) | \`inspect_element\`, \`verify_security\`, \`verify_licenses\` | Non-invasion of \`src/\`, gate preservation, cross-workflow consistency, license compliance | Pipeline credential leakage, supply-chain/action injection, cold-start outage vectors | \`agent-devops\` |
19376
+
19377
+ ## Finding Classifications & Guardrails
19378
+ - **\`agent-auditor\` Labels**: \`[BLOCKING_DEFECT]\`, \`[SIMPLIFICATION]\`, \`[ADVISORY]\`.
19379
+ - **\`agent-adversary\` Proof-of-Break**: (1) Target Entity ID & SHA-256, (2) Failure Scenario / Attack Vector, (3) Concrete PoC, (4) Expected Impact & CVSS v3.1 score.
19380
+ - **Non-Invasive & Anti-Ping-Pong**: Neither auditor nor adversary edits target files directly. Conclude with a canonical Workflow Handoff (\`templates/workflow/agent-handoff.template.md\`) bounded to at most 1 remediation round.
19357
19381
  `
19358
19382
  }
19359
19383
  ]
@@ -19372,9 +19396,14 @@ var SPECIALIZED_AGENT_TEMPLATES = [
19372
19396
  - **Role**: Product Analyst Agent (\`agent-product-analyst\`) and Technical Scribe (AI as Scribe) of the AI-SDLC framework.
19373
19397
  - **Mission**: Capture new product intent via NAF v4 Enterprise & Capability Vision (\`ARCH-INTRO-*\`) and transform natural language business intent into canonical ProductShape artifacts (\`ACT-*\`, \`JRN-*\`, \`UC-*\`, \`BR-*\`, \`TERM-*\`, \`FR-*\`, \`QR-*\`) 100% compliant with JSON schemas.
19374
19398
 
19375
- ## Phase 0: Greenfield Product Inception (New Product Intent)
19376
- - **Trigger**: When the user expresses intent to create a new product from scratch with the framework (or when \`specs/product/\` has no baseline product model yet), initiate **Phase 0** before any SDD change (\`change:new\`) or implementation:
19377
- 1. **Iterative Discovery Interview (Up to 15 Questions or Until Product is Crystal Clear)**: Conduct an iterative discovery interview (up to 15 questions, grouped logically in conversational rounds until zero product ambiguities remain) before drafting specs, clarifying: (a) core problem and mission statement, (b) primary actors/stakeholders (\`ACT-*\`), authentication/role model, and **target user interface modality** (e.g., Web UI, Desktop GUI, Console/TUI/CLI, Mobile app, or headless API/SDK), (c) **modern visual aesthetic & styling standard** (Bootstrap 5-style clean responsive web design with cards/tables/badges, Fluent/modern desktop aesthetic with rounded corners/elevated panels, or colorized rich terminal console with ANSI palettes/status badges), (d) **delivery & operation model** (in plain business terms: Cloud SaaS, Self-Hosted/On-Premise, Local Desktop/CLI binary, or Library/SDK package, plus any data-residency or SLA constraints), (e) **User Journeys (\`JRN-*\`) to generate** \u2014 asking in plain, non-technical language which end-to-end user workflows to model and explicitly guiding the user on the 5 pieces of information needed to build each journey properly: *(1) Who is the protagonist user (\`persona\` / \`ACT-*\`) and their main goal, (2) What are the step-by-step stages from start to finish (\`stages\`), (3) Which screens, channels, or interfaces the user interacts with at each step (\`touchpoints\`), (4) What current obstacles, frustrations, or risks they face (\`pain-points\`), and (5) Which concrete actions or capabilities the system must provide along the journey (\`related-use-cases\` / \`UC-*\`)*, (f) foundational master entities/catalogs and CRUD flows required to operate the system, (g) essential MVP capabilities/use cases (\`UC-*\`) and business rules/invariants (\`BR-*\`) vs. future roadmap, and (h) top 3\u20135 priority quality goals and constraints (\`QR-*\`, \`CON-*\`). Do not proceed to draft artifacts until the product scope is completely clear.
19399
+ ## Phase 0: Greenfield Product Inception & Incremental Discovery (\`/grill-me\` Socratic Decision-Tree Interview Protocol)
19400
+ - **Trigger**: When the user expresses intent to create a new product from scratch with the framework (or when \`specs/product/\` has no baseline product model yet), or when eliciting requirements for a new feature increment with open design decisions, initiate a **\`/grill-me\` Socratic Decision-Tree Interview** before any SDD change (\`change:new\`) or spec generation:
19401
+ 1. **Socratic Decision-Tree Discovery Interview (\`/grill-me\` Mode \u2014 Up to 15 Questions or Until Product is Crystal Clear)**: Instead of dumping a static multi-question form, interview the user relentlessly like \`/grill-me\` by walking down the decision tree **one branch at a time (1\u20132 focused questions per turn, up to 15 questions total)** until zero product ambiguities remain:
19402
+ - **Explore Codebase First (Read-First Rule)**: If any question can be answered by exploring the existing repository (\`package.json\`, \`specs/\`, existing architecture, or code), inspect the codebase first\u2014never ask the user what the repository already answers.
19403
+ - **Opinionated \`(Recommended)\` Choices & Native \`ask_question\` UI**: For every design question, provide concrete options with explicit trade-offs and place your recommended best practice first prefixed with \`(Recommended)\`. Whenever the \`ask_question\` tool is available, use it to present interactive selectable options rather than walls of plain text.
19404
+ - **Active Disambiguation & Constructive Pushback**: Challenge vague answers, unstated assumptions, concurrency/permission conflicts, failure recovery, and scope creep before moving to the next branch.
19405
+ - **Decision-Tree Branches to Resolve**: (a) core problem and mission statement, (b) primary actors/stakeholders (\`ACT-*\`), authentication/role model, and **target user interface modality** (e.g., Web UI, Desktop GUI, Console/TUI/CLI, Mobile app, or headless API/SDK), (c) **modern visual aesthetic & styling standard** (Bootstrap 5-style clean responsive web design with cards/tables/badges, Fluent/modern desktop aesthetic with rounded corners/elevated panels, or colorized rich terminal console with ANSI palettes/status badges), (d) **delivery & operation model** (in plain business terms: Cloud SaaS, Self-Hosted/On-Premise, Local Desktop/CLI binary, or Library/SDK package, plus any data-residency or SLA constraints), (e) **User Journeys (\`JRN-*\`) to generate** \u2014 asking in plain, non-technical language which end-to-end user workflows to model and explicitly guiding the user on the 5 pieces of information needed to build each journey properly: *(1) Who is the protagonist user (\`persona\` / \`ACT-*\`) and their main goal, (2) What are the step-by-step stages from start to finish (\`stages\`), (3) Which screens, channels, or interfaces the user interacts with at each step (\`touchpoints\`), (4) What current obstacles, frustrations, or risks they face (\`pain-points\`), and (5) Which concrete actions or capabilities the system must provide along the journey (\`related-use-cases\` / \`UC-*\`)*, (f) foundational master entities/catalogs and CRUD flows required to operate the system, (g) essential MVP capabilities/use cases (\`UC-*\`) and business rules/invariants (\`BR-*\`) vs. future roadmap, and (h) top 3\u20135 priority quality goals and constraints (\`QR-*\`, \`CON-*\`).
19406
+ - **No-Premature-Drafting Gate**: Strictly forbidden from invoking \`write_file\` or \`edit_file\` to draft specs while decision branches remain unresolved. Once all branches are resolved, present a concise **Decision Blueprint Summary** and confirm with the user before generating files.
19378
19407
  2. **Enterprise & Capability Vision (\`ARCH-INTRO-001\`)**: Draft the root vision artifact (\`type: architecture-introduction\`, \`naf-perspective: "Enterprise & Capability"\`, \`status: draft\`) using \`templates/architecture/introduction-and-goals.template.md\` (\`schemas/architecture/introduction-and-goals.schema.json\`), explicitly recording the agreed UI/interaction modality.
19379
19408
  3. **ProductShape Derivation**: Derive the initial draft artifacts (\`ACT-*\`, \`JRN-*\`, \`UC-*\`, \`BR-*\`, \`FR-*\`, \`QR-*\` with \`status: draft\`) in \`specs/product/\` linked to \`ARCH-INTRO-001\` for human Product Owner review.
19380
19409
  4. **Mandatory End-to-End Presentation Interface & Master Data Prerequisite Rule (ANTI-LIBRARY-ONLY MVP)**: Whenever primary actors (\`ACT-*\`) are human users, the MVP ProductShape baseline MUST include: (a) an explicit presentation/interaction requirement (\`FR-*\`) or vertical UI/CLI acceptance criteria matching the modality agreed in Step 1 (Web UI, Desktop GUI, or Console/TUI/CLI \u2014 never proposing a library-only/in-memory MVP without a runnable entry point), (b) user authentication/session access when multi-role actors exist, and (c) CRUD/configuration flows for foundational master entities (e.g., rooms, staff, students, catalogs) required to operate the core business flows without manual database intervention.
@@ -19391,7 +19420,7 @@ var SPECIALIZED_AGENT_TEMPLATES = [
19391
19420
  - Always write relationships in their single canonical direction (\`derives-from: [UC-*]\`, \`primary-actor: ACT-*\`).
19392
19421
  - In \`UC-*\`, format section \`1. Intent and Outcome\` under the standard English structure: \`As a <ACT-ID>... I want <action>... To <outcome>...\`.
19393
19422
  - In every \`FR-*\` requirement, include a \`\`\`gherkin block with tags \`@<ID> @automated @regression\`, nominal \`Scenario\`, and at least one \`Scenario Outline\` with an \`Examples\` table.
19394
- - Do not assume business rules. If you detect an ambiguity, record it explicitly under \`open-questions\`.
19423
+ - Do not assume business rules. If you detect an ambiguity, probe it via the \`/grill-me\` decision-tree interview or record it explicitly under \`open-questions\`.
19395
19424
  - Run \`pnpm run verify:schemas\` and \`pnpm run verify:duplicates\` before delivering the draft.
19396
19425
  - If autonomy is $\\ge$ \`HUMAN_REVIEW_PLAN\`, conclude by emitting the canonical Workflow Handoff block (\`templates/workflow/agent-handoff.template.md\`) suggesting \`agent-threat-modeler\` or \`agent-expert-user\`, opening the Human Action Window. In \`AUTONOMOUS\` mode, omit interactive handoff.`
19397
19426
  },
@@ -19712,7 +19741,7 @@ var LIFECYCLE_SKILL_TEMPLATES = [
19712
19741
  description: "Orchestrate end-to-end AI-SDLC onboarding and Phase 0 Inception for brand-new (Greenfield) projects, from repository initialization (aisdlc init) and NAF v4 Vision (ARCH-INTRO-*) through ProductShape, STRIDE Threat Modeling, arc42 Architecture, 2-checkpoint transversal auditor/adversary verification, and topological wave SDD delivery.",
19713
19742
  references: "[`process/02_product_definition.md`](../../process/02_product_definition.md), [`process/04b_data_protection_and_privacy.md`](../../process/04b_data_protection_and_privacy.md), [`process/05_architecture_arc42_nafv4.md`](../../process/05_architecture_arc42_nafv4.md), [`process/06_spec_driven_development.md`](../../process/06_spec_driven_development.md), and [`process/09_agent_protocols.md`](../../process/09_agent_protocols.md).",
19714
19743
  globs: "product/**, specs/product/**, architecture/**, templates/**",
19715
- body: '## Purpose & Trigger\nUse this skill when starting a **brand-new software project from scratch (Greenfield)** or when `specs/product/` has no canonical baseline yet. It coordinates the sequential execution of AI-SDLC MCP/CLI tools and specialized role agents so no code is written before product vision, security enclaves, privacy-by-design controls, and architecture contracts exist.\n\n## Consolidated Inception Quality & Adversarial Checkpoint\nAcross Phase 0 and Inception, quality auditing and adversarial stress-testing are concentrated at the completion of Inception design (or after product shaping):\n- **Concurrent Inspection**: Concurrently invoke [`agent-auditor`](../agent-auditor/SKILL.md) (evaluating YAGNI, structural debt, and simplification) and [`agent-adversary`](../agent-adversary/SKILL.md) (stress-testing logic traps, SPOFs, attack vectors, or broken invariants) in clean, isolated context windows.\n- **Non-Invasive Review**: Neither auditor nor adversary modifies deliverables directly; both deliver structured reports citing canonical IDs and SHA-256 digests.\n- **Consolidated Remediation**: Consolidate findings and emit the Workflow Handoff back to authoring agents to resolve blocking defects and apply simplifications (max 2 correction rounds).\n- **Anti-Ping-Pong Guardrail**: Bounded strictly to maximum **1 correction round**. If non-blocking advisories remain, proceed to the downstream phase; if blocking defects persist, escalate immediately to the Human Tech Lead.\n\n## End-to-End Execution Playbook\n\n### Step 1: Branch-Before-Change, Pre-Change Draft PR & Repository Initialization\n1. **Branch-Before-Change & Pre-Change Draft PR/MR**: Verify the active branch (`git branch --show-current`). If currently on a protected branch (`main`, `master`, `release/*`), create and switch to an inception branch (e.g., `feat/inception-baseline` or `chore/aisdlc-inception`) **before** modifying or creating any files. Whenever a remote repository is configured (`git remote`), check if an open PR/MR exists for the branch (`gh pr view` / `gh pr list --head <branch>` on GitHub; `glab mr view` / `glab mr list --source-branch <branch>` on GitLab); if none exists, create an empty bootstrap commit (`git commit --allow-empty -m "chore(inception): open draft PR"`), push (`git push -u origin <branch>`), and open a **Draft PR/MR** (`gh pr create --draft` on GitHub; `glab mr create --draft` or `git push -o merge_request.create -o merge_request.draft` on GitLab) before making any changes (if no remote exists, work locally).\n2. If the workspace has not been initialized with AI-SDLC schemas, policies, and templates, run MCP tool `new` (or CLI `npx @aisdlc/cli init`).\n3. Install the Git telemetry hook via `npx aisdlc git hook install` (`pnpm run git:hook:install`).\n4. Verify `quality-policy.yaml`, `license-policy.yaml`, and `privacy-policy.yaml` are present in the repository root.\n5. Commit and push (`git push`) Phase 0 Inception artifacts (`ARCH-INTRO-001`, ProductShape, Threat/Privacy, Architecture) incrementally to the open Draft PR.\n\n### Step 2: Phase 0 Product Inception (`agent-product-analyst`)\n1. Delegate to or adopt [`agent-product-analyst`](../agent-product-analyst/SKILL.md).\n2. **Iterative Discovery Interview (Up to 15 Questions or Until Product Scope is 100% Clear)**: Before generating artifacts, clarify with the human Product Owner in rounds of 3\u20135 questions:\n - Core mission statement and business problem.\n - Primary actors (`ACT-*`) and external systems.\n - **Target Interaction & Presentation Modality (`CON-*` / `ARCH-INTRO-001`)**: Explicitly define at inception how human actors will interact with the MVP (e.g., **Web UI**, **Desktop GUI**, **Interactive Console/TUI/CLI**, **Mobile App**, or **Headless API/SDK** if actors are exclusively external systems). Never assume a headless class library when human actors exist.\n - **Delivery & Deployment Model (`ARCH-INTRO-001` / `CON-*` / `QR-*`)**: Clarify in business terms how the product will be delivered and operated (Cloud SaaS, Self-Hosted/On-Premise, Local Desktop/CLI binary, or Library/SDK package) along with any data-residency or availability requirements.\n - **Modern Aesthetic & Design System Standard (`CON-UI-*` / `ARCH-INTRO-001`)**: Explicitly define a modern, visually attractive styling standard adapted to the chosen presentation modality: (1) **Web UI**: modern responsive layouts using Bootstrap 5 utility classes and components (`card`, `table table-hover`, `badge`, `alert alert-*`, responsive grid) or modern equivalents, (2) **Desktop GUI**: modern clean design (rounded corners, elevated panels, padding, consistent typography) avoiding raw legacy unstyled windows, (3) **Console / TUI**: rich terminal interface with ANSI colors, status badges, formatted tables, and progress indicators (e.g., Spectre.Console, Rich, Ink, or Chalk/Picocolors).\n - **User Journeys (`JRN-*`) to Generate (Plain-Language Elicitation)**: Ask in simple, everyday language which end-to-end user journeys to model, explicitly indicating the 5 pieces of information needed to build each journey properly: (1) protagonist user (`persona` / `ACT-*`) and their main goal, (2) step-by-step stages from start to finish (`stages`), (3) interaction screens/channels at each stage (`touchpoints`), (4) current obstacles or pain points (`pain-points`), and (5) concrete system capabilities/use cases (`related-use-cases` / `UC-*`).\n - Essential MVP capabilities (`UC-*`) vs. future roadmap, ensuring **foundational entity CRUD / master data workflows** are explicitly included in the MVP.\n - Top 3\u20135 quality goals (`QR-*`) and regulatory/technical constraints (`CON-*`).\n3. **Enterprise & Capability Vision**: Create `specs/architecture/ARCH-INTRO-001.md` (`status: draft`, `naf-perspective: "Enterprise & Capability"`) using `templates/architecture/introduction-and-goals.template.md`, explicitly recording the chosen presentation/UI modality.\n4. **ProductShape Baseline**: Derive initial draft artifacts in `specs/product/`:\n - Actors (`ACT-*`), User Journeys (`JRN-*`), Use Cases (`UC-*`), Business Rules (`BR-*`), Functional Requirements (`FR-*` with mandatory `@nominal`, `@boundary`, `@invalid` Gherkin blocks), and Quality Requirements (`QR-*`).\n - **Vertical Slice Completeness Rule**: Every human-triggered `UC-*` in the MVP must specify observable user interaction through the agreed presentation modality and state persistence.\n5. Validate schemas and duplicates via MCP tools `verify_schemas` and `verify_duplicates` (or `npx aisdlc check --fix`).\n6. **Mandatory Upstream UX & MVP Boundary Gate (`agent-expert-user`)**: Invoke [`agent-expert-user`](../agent-expert-user/SKILL.md) in **Design Mode** to stress-test the proposed workflows, enforce strict MVP boundaries, reject any headless/library-only MVP when human actors exist, verify master-data CRUD prerequisites, and separate core MVP `UC-*` from future roadmap backlog.\n\n### Step 3: Shift-Left Cybersecurity & Privacy-by-Design (`agent-threat-modeler` \u2794 `agent-privacy-officer`)\n1. Delegate to or adopt [`agent-threat-modeler`](../agent-threat-modeler/SKILL.md).\n2. Analyze every `UC-*` and `FR-*` using STRIDE and OWASP ASVS to create in `specs/security/`:\n - Threat Actors (`ACT-THREAT-*`), Abuse Cases (`ABUSE-*` targeting `UC-*`), and Security Requirements (`SEC-REQ-*` with `@security @mitigation` negative Gherkin scenarios and `SEC-ENC-*` Zero Trust enclave bindings).\n3. Hand off to [`agent-privacy-officer`](../agent-privacy-officer/SKILL.md) to apply LINDDUN and GDPR/LOPDGDD/ISO 27701 in `specs/privacy/`:\n - Personal Data Assets (`DATA-ASSET-*`), GDPR Art. 30 Records of Processing (`PRIV-ROPA-*`), and Privacy Requirements (`PRIV-REQ-*` with `@privacy @gdpr` Gherkin scenarios).\n4. Run `verify_schemas` and `verify_privacy` (`npx aisdlc verify privacy`).\n\n### Step 4: Modular Systems Architecture (`agent-system-architect`)\n1. Delegate to or adopt [`agent-system-architect`](../agent-system-architect/SKILL.md).\n2. Define building blocks (`CMP-*`), Deployment Topology (`ARCH-DEP-001` translating the delivery model into execution nodes `RES-*` and enclaves `SEC-ENC-*`), and Architectural Decision Records (`ADR-*`) in `specs/architecture/` using `templates/architecture/`, ensuring every `CMP-*` cites `implements-use-cases` (`UC-*`) and `satisfies-requirements` (`FR-*`, `QR-*`, `SEC-REQ-*`, `PRIV-REQ-*`).\n3. Inspect third-party stack choices against `license-policy.yaml` via `verify_licenses` and subprocessors against `privacy-policy.yaml` via `verify_privacy`.\n4. **Technical Security & Privacy Feedback Loop**: If new infrastructure, persistence, LLM, or auth decisions introduce attack vectors or PII flows, return to `agent-threat-modeler` \u2794 `agent-privacy-officer` to add technical mitigations before coding.\n\n### Checkpoint 1 (Inception Design Checkpoint): `agent-auditor` $\\parallel$ `agent-adversary`\n1. Concurrently invoke `agent-auditor` (evaluating YAGNI, architecture bloat, interface contracts) and `agent-adversary` (simulating SPOFs, race conditions, unhandled threat scenarios) in isolated context windows over all Phase 0 deliverables.\n2. Consolidate findings and remediate in authoring agents (max 2 correction rounds), committing and pushing (`git push`) updates to the Draft PR.\n3. **Human Approval Gate**: Mark the Inception Draft PR/MR ready (`gh pr ready` / `glab mr update --ready`) and present the comprehensive Inception model for Human PO & Tech Lead sign-off.\n\n### Step 5: Release Milestone & Topological Wave SDD Plan (`1 UC-* = 1 CHG-*`)\n1. **Open Release Milestone (`Tier 2: release/v1.0.0` or Streamlined `main`)**: Group approved MVP Use Cases (`UC-001`, `UC-002`, ...) and visualize execution plan.\n2. **Decompose into Topological Waves (DAG Parallelism)**:\n - **Wave 0 \u2014 Foundational Core, Executable Presentation Host & CI/CD (`agent-devops`) (Sequential)**: Scaffold `CHG-001` via [`aisdlc-increment`](../aisdlc-increment/SKILL.md) to establish domain core, shared contracts, auth/config, and the runnable presentation host. Invoke [`agent-devops`](../agent-devops/SKILL.md) to clarify the technical CI/CD & packaging strategy (CI Quality Gates, Docker/GHCR, package registry, or PaaS/Cloud) and scaffold `.github/workflows/` and container manifests without modifying `src/`.\n - **Waves 1..N \u2014 Orthogonal End-to-End Vertical Slices (Parallel Multi-Agent)**: Once Wave 0 contracts are merged, execute independent `CHG-*` increments (`1 UC-* = 1 CHG-*`) in parallel across isolated branches/worktrees (`feat/CHG-002`, `feat/CHG-003`, ...).\n3. **Release Consolidation (`agent-devops`)**: Once all MVP increments are integrated (`sdd_integrate`), run `verify` (`npx aisdlc verify all`), generate SBOM, compute release KPIs (`kpi_release`), regenerate canonical reports and dashboard (`report_quality`, `report_dashboard`), verify release/publishing workflows via `agent-devops`, and open the Release PR for final human approval and SemVer tagging (`v1.0.0`). Individual `feat/CHG-*` PRs must compute reports locally without committing global `reports/` files to avoid merge conflicts.\n\n## Non-Negotiable Guardrails\n- **Clean Context Window Isolation per Specialist Agent (Cross-IDE)**: Never execute two distinct specialist roles in the same conversation context window. Always spawn each specialist in a fresh isolated subagent context that reads state strictly from disk.\n- **2-Checkpoint Quality & Adversarial Protocol**: Quality auditing and adversarial testing are concentrated at the completion of Inception design and Pre-PR implementation, avoiding micro-step overhead.\n- **FORBIDDEN Library-Only MVP for Human-Facing Products**: Whenever a product defines human actors (`ACT-*`), no agent may scope the MVP as headless class libraries without an executable presentation host.\n- **FORBIDDEN Greenfield Monolith (`1 UC-* = 1 CHG-*`)**: Never bundle the entire product or multiple `UC-*` into a single SDD change (`CHG-001`). Parallelism is achieved by running **multiple bounded `CHG-*` concurrently** on separate branches/worktrees.\n- Never invoke `sdd_change_new` or write application code in `src/` on a Greenfield project before `ARCH-INTRO-001`, `agent-expert-user` MVP validation, and at least one `UC-*` and `FR-*` exist.\n- All newly drafted specifications must start with `status: draft`.'
19744
+ body: '## Purpose & Trigger\nUse this skill when starting a **brand-new software project from scratch (Greenfield)** or when `specs/product/` has no canonical baseline yet. It coordinates the sequential execution of AI-SDLC MCP/CLI tools and specialized role agents so no code is written before product vision, security enclaves, privacy-by-design controls, and architecture contracts exist.\n\n## Consolidated Inception Quality & Adversarial Checkpoint\nAcross Phase 0 and Inception, quality auditing and adversarial stress-testing are concentrated at the completion of Inception design (or after product shaping):\n- **Concurrent Inspection**: Concurrently invoke [`agent-auditor`](../agent-auditor/SKILL.md) (evaluating YAGNI, structural debt, and simplification) and [`agent-adversary`](../agent-adversary/SKILL.md) (stress-testing logic traps, SPOFs, attack vectors, or broken invariants) in clean, isolated context windows.\n- **Non-Invasive Review**: Neither auditor nor adversary modifies deliverables directly; both deliver structured reports citing canonical IDs and SHA-256 digests.\n- **Consolidated Remediation**: Consolidate findings and emit the Workflow Handoff back to authoring agents to resolve blocking defects and apply simplifications (max 2 correction rounds).\n- **Anti-Ping-Pong Guardrail**: Bounded strictly to maximum **1 correction round**. If non-blocking advisories remain, proceed to the downstream phase; if blocking defects persist, escalate immediately to the Human Tech Lead.\n\n## End-to-End Execution Playbook\n\n### Step 1: Branch-Before-Change, Pre-Change Draft PR & Repository Initialization\n1. **Branch-Before-Change & Pre-Change Draft PR/MR**: Verify the active branch (`git branch --show-current`). If currently on a protected branch (`main`, `master`, `release/*`), create and switch to an inception branch (e.g., `feat/inception-baseline` or `chore/aisdlc-inception`) **before** modifying or creating any files. Whenever a remote repository is configured (`git remote`), check if an open PR/MR exists for the branch (`gh pr view` / `gh pr list --head <branch>` on GitHub; `glab mr view` / `glab mr list --source-branch <branch>` on GitLab); if none exists, create an empty bootstrap commit (`git commit --allow-empty -m "chore(inception): open draft PR"`), push (`git push -u origin <branch>`), and open a **Draft PR/MR** (`gh pr create --draft` on GitHub; `glab mr create --draft` or `git push -o merge_request.create -o merge_request.draft` on GitLab) before making any changes (if no remote exists, work locally).\n2. If the workspace has not been initialized with AI-SDLC schemas, policies, and templates, run MCP tool `new` (or CLI `npx @aisdlc/cli init`).\n3. Install the Git telemetry hook via `npx aisdlc git hook install` (`pnpm run git:hook:install`).\n4. Verify `quality-policy.yaml`, `license-policy.yaml`, and `privacy-policy.yaml` are present in the repository root.\n5. Commit and push (`git push`) Phase 0 Inception artifacts (`ARCH-INTRO-001`, ProductShape, Threat/Privacy, Architecture) incrementally to the open Draft PR.\n\n### Step 2: Phase 0 Product Inception (`agent-product-analyst` \u2014 `/grill-me` Socratic Decision-Tree Interview)\n1. Delegate to or adopt [`agent-product-analyst`](../agent-product-analyst/SKILL.md).\n2. **Socratic Decision-Tree Discovery Interview (`/grill-me` Mode \u2014 Up to 15 Questions or Until Product Scope is 100% Clear)**: Before generating artifacts, interview the human Product Owner relentlessly like `/grill-me` by walking down the decision tree **one branch at a time (1\u20132 focused questions per turn, up to 15 questions total)**, exploring the repository first (Read-First Rule), presenting opinionated `(Recommended)` options via `ask_question` when available, challenging vague assumptions/edge cases, and enforcing the **No-Premature-Drafting Gate** (no spec files written until all branches are resolved and the user confirms the Decision Blueprint Summary):\n - Core mission statement and business problem.\n - Primary actors (`ACT-*`) and external systems.\n - **Target Interaction & Presentation Modality (`CON-*` / `ARCH-INTRO-001`)**: Explicitly define at inception how human actors will interact with the MVP (e.g., **Web UI**, **Desktop GUI**, **Interactive Console/TUI/CLI**, **Mobile App**, or **Headless API/SDK** if actors are exclusively external systems). Never assume a headless class library when human actors exist.\n - **Delivery & Deployment Model (`ARCH-INTRO-001` / `CON-*` / `QR-*`)**: Clarify in business terms how the product will be delivered and operated (Cloud SaaS, Self-Hosted/On-Premise, Local Desktop/CLI binary, or Library/SDK package) along with any data-residency or availability requirements.\n - **Modern Aesthetic & Design System Standard (`CON-UI-*` / `ARCH-INTRO-001`)**: Explicitly define a modern, visually attractive styling standard adapted to the chosen presentation modality: (1) **Web UI**: modern responsive layouts using Bootstrap 5 utility classes and components (`card`, `table table-hover`, `badge`, `alert alert-*`, responsive grid) or modern equivalents, (2) **Desktop GUI**: modern clean design (rounded corners, elevated panels, padding, consistent typography) avoiding raw legacy unstyled windows, (3) **Console / TUI**: rich terminal interface with ANSI colors, status badges, formatted tables, and progress indicators (e.g., Spectre.Console, Rich, Ink, or Chalk/Picocolors).\n - **User Journeys (`JRN-*`) to Generate (Plain-Language Elicitation)**: Ask in simple, everyday language which end-to-end user journeys to model, explicitly indicating the 5 pieces of information needed to build each journey properly: (1) protagonist user (`persona` / `ACT-*`) and their main goal, (2) step-by-step stages from start to finish (`stages`), (3) interaction screens/channels at each stage (`touchpoints`), (4) current obstacles or pain points (`pain-points`), and (5) concrete system capabilities/use cases (`related-use-cases` / `UC-*`).\n - Essential MVP capabilities (`UC-*`) vs. future roadmap, ensuring **foundational entity CRUD / master data workflows** are explicitly included in the MVP.\n - Top 3\u20135 quality goals (`QR-*`) and regulatory/technical constraints (`CON-*`).\n3. **Enterprise & Capability Vision**: Create `specs/architecture/ARCH-INTRO-001.md` (`status: draft`, `naf-perspective: "Enterprise & Capability"`) using `templates/architecture/introduction-and-goals.template.md`, explicitly recording the chosen presentation/UI modality.\n4. **ProductShape Baseline**: Derive initial draft artifacts in `specs/product/`:\n - Actors (`ACT-*`), User Journeys (`JRN-*`), Use Cases (`UC-*`), Business Rules (`BR-*`), Functional Requirements (`FR-*` with mandatory `@nominal`, `@boundary`, `@invalid` Gherkin blocks), and Quality Requirements (`QR-*`).\n - **Vertical Slice Completeness Rule**: Every human-triggered `UC-*` in the MVP must specify observable user interaction through the agreed presentation modality and state persistence.\n5. Validate schemas and duplicates via MCP tools `verify_schemas` and `verify_duplicates` (or `npx aisdlc check --fix`).\n6. **Mandatory Upstream UX & MVP Boundary Gate (`agent-expert-user`)**: Invoke [`agent-expert-user`](../agent-expert-user/SKILL.md) in **Design Mode** to stress-test the proposed workflows, enforce strict MVP boundaries, reject any headless/library-only MVP when human actors exist, verify master-data CRUD prerequisites, and separate core MVP `UC-*` from future roadmap backlog.\n\n### Step 3: Shift-Left Cybersecurity & Privacy-by-Design (`agent-threat-modeler` \u2794 `agent-privacy-officer`)\n1. Delegate to or adopt [`agent-threat-modeler`](../agent-threat-modeler/SKILL.md).\n2. Analyze every `UC-*` and `FR-*` using STRIDE and OWASP ASVS to create in `specs/security/`:\n - Threat Actors (`ACT-THREAT-*`), Abuse Cases (`ABUSE-*` targeting `UC-*`), and Security Requirements (`SEC-REQ-*` with `@security @mitigation` negative Gherkin scenarios and `SEC-ENC-*` Zero Trust enclave bindings).\n3. Hand off to [`agent-privacy-officer`](../agent-privacy-officer/SKILL.md) to apply LINDDUN and GDPR/LOPDGDD/ISO 27701 in `specs/privacy/`:\n - Personal Data Assets (`DATA-ASSET-*`), GDPR Art. 30 Records of Processing (`PRIV-ROPA-*`), and Privacy Requirements (`PRIV-REQ-*` with `@privacy @gdpr` Gherkin scenarios).\n4. Run `verify_schemas` and `verify_privacy` (`npx aisdlc verify privacy`).\n\n### Step 4: Modular Systems Architecture (`agent-system-architect`)\n1. Delegate to or adopt [`agent-system-architect`](../agent-system-architect/SKILL.md).\n2. Define building blocks (`CMP-*`), Deployment Topology (`ARCH-DEP-001` translating the delivery model into execution nodes `RES-*` and enclaves `SEC-ENC-*`), and Architectural Decision Records (`ADR-*`) in `specs/architecture/` using `templates/architecture/`, ensuring every `CMP-*` cites `implements-use-cases` (`UC-*`) and `satisfies-requirements` (`FR-*`, `QR-*`, `SEC-REQ-*`, `PRIV-REQ-*`).\n3. Inspect third-party stack choices against `license-policy.yaml` via `verify_licenses` and subprocessors against `privacy-policy.yaml` via `verify_privacy`.\n4. **Technical Security & Privacy Feedback Loop**: If new infrastructure, persistence, LLM, or auth decisions introduce attack vectors or PII flows, return to `agent-threat-modeler` \u2794 `agent-privacy-officer` to add technical mitigations before coding.\n\n### Checkpoint 1 (Inception Design Checkpoint): `agent-auditor` $\\parallel$ `agent-adversary`\n1. Concurrently invoke `agent-auditor` (evaluating YAGNI, architecture bloat, interface contracts) and `agent-adversary` (simulating SPOFs, race conditions, unhandled threat scenarios) in isolated context windows over all Phase 0 deliverables.\n2. Consolidate findings and remediate in authoring agents (max 2 correction rounds), committing and pushing (`git push`) updates to the Draft PR.\n3. **Human Approval Gate**: Mark the Inception Draft PR/MR ready (`gh pr ready` / `glab mr update --ready`) and present the comprehensive Inception model for Human PO & Tech Lead sign-off.\n\n### Step 5: Release Milestone & Topological Wave SDD Plan (`1 UC-* = 1 CHG-*`)\n1. **Open Release Milestone (`Tier 2: release/v1.0.0` or Streamlined `main`)**: Group approved MVP Use Cases (`UC-001`, `UC-002`, ...) and visualize execution plan.\n2. **Decompose into Topological Waves (DAG Parallelism)**:\n - **Wave 0 \u2014 Foundational Core, Executable Presentation Host & CI/CD (`agent-devops`) (Sequential)**: Scaffold `CHG-001` via [`aisdlc-increment`](../aisdlc-increment/SKILL.md) to establish domain core, shared contracts, auth/config, and the runnable presentation host. Invoke [`agent-devops`](../agent-devops/SKILL.md) to clarify the technical CI/CD & packaging strategy (CI Quality Gates, Docker/GHCR, package registry, or PaaS/Cloud) and scaffold `.github/workflows/` and container manifests without modifying `src/`.\n - **Waves 1..N \u2014 Orthogonal End-to-End Vertical Slices (Parallel Multi-Agent)**: Once Wave 0 contracts are merged, execute independent `CHG-*` increments (`1 UC-* = 1 CHG-*`) in parallel across isolated branches/worktrees (`feat/CHG-002`, `feat/CHG-003`, ...).\n3. **Release Consolidation (`agent-devops`)**: Once all MVP increments are integrated (`sdd_integrate`), run `verify` (`npx aisdlc verify all`), generate SBOM, compute release KPIs (`kpi_release`), regenerate canonical reports and dashboard (`report_quality`, `report_dashboard`), verify release/publishing workflows via `agent-devops`, and open the Release PR for final human approval and SemVer tagging (`v1.0.0`). Individual `feat/CHG-*` PRs must compute reports locally without committing global `reports/` files to avoid merge conflicts.\n\n## Non-Negotiable Guardrails\n- **Clean Context Window Isolation per Specialist Agent (Cross-IDE)**: Never execute two distinct specialist roles in the same conversation context window. Always spawn each specialist in a fresh isolated subagent context that reads state strictly from disk.\n- **2-Checkpoint Quality & Adversarial Protocol**: Quality auditing and adversarial testing are concentrated at the completion of Inception design and Pre-PR implementation, avoiding micro-step overhead.\n- **FORBIDDEN Library-Only MVP for Human-Facing Products**: Whenever a product defines human actors (`ACT-*`), no agent may scope the MVP as headless class libraries without an executable presentation host.\n- **FORBIDDEN Greenfield Monolith (`1 UC-* = 1 CHG-*`)**: Never bundle the entire product or multiple `UC-*` into a single SDD change (`CHG-001`). Parallelism is achieved by running **multiple bounded `CHG-*` concurrently** on separate branches/worktrees.\n- Never invoke `sdd_change_new` or write application code in `src/` on a Greenfield project before `ARCH-INTRO-001`, `agent-expert-user` MVP validation, and at least one `UC-*` and `FR-*` exist.\n- All newly drafted specifications must start with `status: draft`.'
19716
19745
  },
19717
19746
  {
19718
19747
  id: "aisdlc-legacy-onboarding",
@@ -19728,7 +19757,7 @@ var LIFECYCLE_SKILL_TEMPLATES = [
19728
19757
  description: "Orchestrate the complete Spec-Driven Development (SDD) lifecycle for bounded feature increments, enhancements, and bugfixes (CHG-* / TSK-*), from upstream requirement derivation and streamlined Git branching through Red-Green-Refactor TDD, 2-checkpoint transversal auditor/adversary verification, all 12 Quality Gates, and canonical baseline integration.",
19729
19758
  references: "[`process/06_spec_driven_development.md`](../../process/06_spec_driven_development.md), [`process/08_citation_contract_and_drift.md`](../../process/08_citation_contract_and_drift.md), [`process/10_quality_management_and_release_gates.md`](../../process/10_quality_management_and_release_gates.md), [`process/11_git_branching_and_lifecycle.md`](../../process/11_git_branching_and_lifecycle.md), and [`process/09_agent_protocols.md`](../../process/09_agent_protocols.md).",
19730
19759
  globs: "specs/changes/**, src/**, packages/**, tests/**",
19731
- body: '## Purpose & Trigger\nUse this skill for **every evolutionary increment (new feature, enhancement, refactoring, or bugfix)** on a project that already has an AI-SDLC baseline. It enforces strict 360\xB0 traceability from upstream ProductShape (`UC-*`, `FR-*`) through SDD (`CHG-*`, `HOF-*`, `TSK-*`), TDD implementation, the 12 Quality Gates, and baseline promotion under the **streamlined 2-checkpoint architecture**.\n\n## End-to-End Execution Playbook\n\n### Step 0: Branch-Before-Change, Pre-Change Draft PR & Scope Triage (`patch | standard | critical`)\n1. **Branch-Before-Change & Pre-Change Draft PR/MR**: Verify the active branch (`git branch --show-current`). If currently on a protected branch (`main`, `master`, `release/*`), create and switch to a dedicated working branch (`feat/<CHG-ID>-<slug>` or `fix/<CHG-ID>-<slug>`) **before** modifying or creating any files. Whenever a remote repository is configured (`git remote`), check if an open PR/MR already exists for the branch (`gh pr view` / `gh pr list --head <branch>` on GitHub; `glab mr view` / `glab mr list --source-branch <branch>` on GitLab); if none exists, create an empty bootstrap commit (`git commit --allow-empty -m "chore(<CHG-ID>): open draft PR"`), push (`git push -u origin <branch>`), and open a **Draft PR/MR** (`gh pr create --draft` on GitHub; `glab mr create --draft` or `git push -o merge_request.create -o merge_request.draft` on GitLab) before making any changes (if no remote exists, work locally).\n2. Before editing any files or running `sdd_change_new`, inspect if the user explicitly specified the friction profile (`patch`, `standard`, or `critical`). If not, analyze the requested scope and ask the user via interactive question (`ask_question`) with the recommended option first (`(Recommended)`):\n - **`patch`**: low-risk isolated bugfixes ($\\le 2$ source files, zero Anti-Bypass paths), typos, or linter fixes on `patch/*` or `fix/*` branches (`npx aisdlc change new "<Title>" --id <chg-id> --profile patch` generates `spec.md` with `profile: patch` and `verification` block only, omitting `design.md`, `tasks.md`, and `handoff.yaml`; fast-track directly to Step 3 -> Step 4 -> Step 5 -> Step 6).\n - **`standard`**: functional increments, new/updated `UC-*` or `FR-*`, or contract changes (`proposal.md`, `spec.md`, `design.md`, `tasks.md`, and `handoff.yaml`).\n - **`critical`**: changes touching security enclaves (`SEC-ENC-*`), sensitive PII (`SPECIAL_CATEGORY` / `DPIA-*`), schemas, or migrations (full `standard` artifacts + `ADR-*` + mandatory STRIDE modeling and dual human approval).\n2. **Anti-Bypass Rule**: Never recommend or allow `patch` when the change touches protected paths (`schemas/`, `specs/security/`, `specs/privacy/`, `packages/security/`, `SEC-ENC-*`, `09_decisions/`, `migrations/`, or `*-policy.yaml`).\n\n### Step 1: Upstream Product, Security & Privacy Delta (`agent-product-analyst`, `agent-threat-modeler` & `agent-privacy-officer`)\n1. Inspect active changes via MCP tool `get_active_handoff` and verify whether the target Use Case (`UC-*`) or Requirement (`FR-*`) already exists in `specs/product/`.\n2. If the increment introduces new behavior or modifies business rules:\n - Use [`agent-product-analyst`](../agent-product-analyst/SKILL.md) to create or update the `UC-*`, `BR-*`, and `FR-*` (with `@nominal`, `@boundary`, `@invalid` Gherkin scenarios).\n - Run MCP tool `verify_duplicates` (`npx aisdlc verify duplicates`) to block lexical or functional collisions early.\n - Use [`agent-threat-modeler`](../agent-threat-modeler/SKILL.md) to model any new `ABUSE-*` and `SEC-REQ-*` mitigations.\n - Hand off to [`agent-privacy-officer`](../agent-privacy-officer/SKILL.md) to model any personal data assets (`DATA-ASSET-*`), GDPR Art. 30 processing activities (`PRIV-ROPA-*`), and privacy requirements (`PRIV-REQ-*`).\n3. Run `sdd_check_fix` (`npx aisdlc check --fix`) to sync Gherkin `.feature` files and validate schemas.\n\n### Step 2: SDD Change Scaffolding & Architectural Design (`agent-system-architect`)\n1. Scaffold the bounded change package and its cryptographic `handoff.yaml` sidecar (`HOF-*`) targeting **a single Use Case (`1 UC-* = 1 CHG-*`)** using MCP tool `sdd_change_new` (or CLI `npx aisdlc change new "<Title>" --from <UC-ID> --profile <patch|standard|critical>`).\n2. Use [`agent-system-architect`](../agent-system-architect/SKILL.md) to complete `specs/changes/active/<chg-id>/design.md` and `specs/architecture/CMP-*.md`:\n - Structure `CMP-*` and `design.md` around **Modular & Hexagonal Architecture (Ports & Adapters)** (Domain/Application Core + Inbound/Outbound Ports) and define **Contract-First** shared interfaces/DTOs/schemas when spanning Frontend and Backend.\n - Map 100% of `handoff.yaml` requirements and components to atomic tasks (`TSK-*`) in `tasks.md` with explicit `autonomy-mode` (`AUTONOMOUS`, `HUMAN_REVIEW_PLAN`, `HIGH_RISK_MANUAL`, `AMBIGUOUS`).\n3. Run `verify_governance` and `verify_friction` to validate task governance, 100% design/handoff scope coverage, and anti-monolith limits (`maxUseCasesPerChange = 1`).\n4. Commit the SDD scaffolding and specification/design artifacts on the working branch and push (`git push`) to the open Draft PR.\n\n### Checkpoint 1: Specification & Design Checkpoint (`agent-auditor` $\\parallel$ `agent-adversary`)\n1. Concurrently invoke `agent-auditor` (evaluating YAGNI, interface bloat, and task completeness) and `agent-adversary` (probing contradictions, race conditions, and architecture SPOFs) in clean, isolated context windows over the combined upstream deliverables.\n2. Consolidate findings and remediate any blocking defects (max 2 correction rounds), committing and pushing updates (`git push`) to the open Draft PR.\n3. **Human Approval Gate 1**: Present the specification and architectural design (reviewable directly in the Draft PR) for Human Tech Lead / PO sign-off before proceeding to code.\n\n### Step 3: Streamlined Git Branch Navigation & Multi-Agent Parallelism\n1. Confirm you are on the working branch via `npx aisdlc git checkout <TSK-ID>` (`pnpm run git:checkout <TSK-ID>`).\n - **Default Streamlined Mode**: Ensures the agent is on `feat/<CHG-ID>-<slug>` (already created in Step 0).\n - Granular task traceability is maintained via commit trailers (`Task-ID: TSK-*`, `Parent-Ref: CHG-*`) and commit scopes (`feat(TSK-001): ...`), enabling per-task rework KPIs without branch explosion.\n2. **Parallel Agent Execution Rules (`--parallel`)**:\n - When concurrent multi-agent implementation is configured (e.g. Backend adapter $\\parallel$ Frontend UI), `aisdlc git checkout <TSK-ID> --parallel` creates isolated worktrees (`task/<PARENT-ID>/<TSK-ID>-<slug>`).\n\n### Step 4: Red Phase \u2014 Failing Test Suites (`agent-qa-engineer`)\n1. Delegate to or adopt [`agent-qa-engineer`](../agent-qa-engineer/SKILL.md).\n2. Run `npx aisdlc gherkin extract` (`sdd_check_fix`) to deterministically extract `.feature` files from Markdown specifications, and run `npx aisdlc gherkin scaffold <feature-file> --lang <lang> --fallback` to scaffold RED step definitions in `tests/steps/` (never hand-author `.feature` or step skeletons).\n3. Author FAILING (red) assertions inside the scaffolded step definitions, plus integration/contract tests and mandatory E2E/UI tests (evaluating Happy Path, Sad Path, and Boundary Values for every interactive UI element, plus `UC-*` alternative/exception flows) in `tests/`.\n4. Commit failing test suites with trailer `Task-ID: TSK-YYY` and push (`git push`) to the Draft PR (CI workflows skip heavy checks while the PR is in `draft` state).\n5. Run `npx aisdlc verify gherkin` and `verify_testing` (`npx aisdlc verify testing`) to confirm deterministic Gherkin/step sync and requirement test coverage before writing production code.\n\n### Step 5: Green & Refactor Phase \u2014 Surgical Implementation (`agent-developer` / `agent-devops`)\n1. Delegate application tasks (`src/`, `tests/`) to [`agent-developer`](../agent-developer/SKILL.md). For tasks (`TSK-*`) targeting CI/CD workflows (`.github/workflows/`), `Dockerfile*`, IaC manifests, or support scripts (`scripts/`), delegate to [`agent-devops`](../agent-devops/SKILL.md) under its strict non-invasion guardrail on `src/`.\n2. Execute the **Mandatory 4-Step Implementation Order**:\n - (1) Modular & Hexagonal Architecture (Ports & Adapters) boundaries, (2) Inter-module interfaces/ports first, (3) Concrete component implementation (zero fake-wiring; all 4 UI states `Loading`, `Empty`, `Error`, `Nominal`), (4) Exhaustive per-function unit tests in `tests/unit/` (Happy Path, Sad Path, Boundary, Null/Empty).\n3. Commit implementations with trailer `Task-ID: TSK-YYY` and push (`git push`) incrementally to the Draft PR.\n4. Verify any new dependency against `license-policy.yaml` via `verify_licenses` and PII handling against `privacy-policy.yaml` via `verify_privacy`.\n5. Run `sdd_check_fix` (`npx aisdlc check --fix`) and `verify_quality` (`npx aisdlc verify quality`) until Cyclomatic Complexity $\\le 10$, Cognitive $\\le 15$, Maintainability $\\ge 50$, and Function Length $\\le 40$ lines pass with zero violations.\n6. Run `verify` (`npx aisdlc verify all`) confirming all **12 Quality Gates** pass green.\n\n### Checkpoint 2: Implementation & Pre-PR Checkpoint (`agent-auditor` $\\parallel$ `agent-adversary`)\n1. **Functional Validation (`agent-expert-user`)**: Validate live runtime behavior against `UC-*` and `FR-*` criteria (including `0` browser `console.error`/`pageerror` on UI changes).\n2. Concurrently invoke `agent-auditor` (verifying Clean Code, SOLID, DRY, AST thresholds, and eliminating unnecessary abstractions) and `agent-adversary` (performing conceptual test mutation to verify QA tests catch defects, and attacking boundary inputs with concrete PoC breaks).\n3. Consolidate findings and remediate in `agent-developer` (max 2 correction rounds), committing and pushing (`git push`) fixes to the Draft PR.\n4. Run `npx aisdlc verify all` to confirm final green status, then run `report_quality` (`npx aisdlc report quality`) and `report_dashboard` (`npx aisdlc report dashboard`) to compute local reports and the interactive dashboard for review (**do NOT stage or commit global `reports/*` files on `feat/CHG-*` branches** to prevent merge conflicts across parallel PRs).\n\n### Step 6: Human PR Sign-Off, Baseline Integration & Release Closure\n1. **Ready for Review & Feature Merge (`feat/CHG-*` \u2794 `main` or `release/vX.Y.Z`)**:\n - Update the Draft PR body with the Plan Execution Matrix (`[X]` vs `[O]`), KPI summary (`kpi_pr` including per-task rework metrics), and local verification/report summary (`reports/QUALITY_REPORT.md`, `reports/TRACEABILITY_MATRIX.md`, `reports/dashboard.html` inspected locally, without committing global `reports/*` to the feature PR), then run `gh pr ready` (GitHub) or `glab mr update --ready` (GitLab) to mark the PR/MR **Ready for Review** (triggering full CI verification) and present it to the human Tech Lead for review and approval (never auto-approve or auto-merge).\n2. **Canonical Baseline Promotion**: Once tasks in `tasks.md` are marked `COMPLETED` and approved, promote the change into the canonical baseline via MCP tool `sdd_integrate` (`npx aisdlc sdd integrate --auto`), which synchronizes the normative text and BDD Gherkin scenarios into `FR-*` / `UC-*` / `SEC-REQ-*` / `PRIV-REQ-*`, bumps minor SemVer when updating active specs, resyncs PDaC SHA-256 digests, promotes `FR-*` to `active`, links `CMP-*`, and archives the change under `specs/changes/completed/` (or `npx aisdlc sdd integrate --all` to consolidate all completed changes).\n3. **Semantic Versioning Promotion**:\n - **Profile-Driven SemVer Mapping**: Bump the version of the software under development according to the increment\'s friction profile:\n - `patch` -> bump **Patch** digit (`x.y.Z`)\n - `standard` -> bump **Minor** digit (`x.Y.0`)\n - `critical` -> bump **Major** digit (`X.0.0`)\n - **Highest-Precedence Consolidation (Multiple Pending Increments)**: When multiple pending increments exist in the active branch or release batch being integrated, apply **ONLY a single version bump corresponding to the highest rank** (`critical` / Major > `standard` / Minor > `patch` / Patch), ignoring unrelated historical directories.\n - Tag the merge with the computed SemVer bump (`patch`, `minor`, `major`) and refresh the interactive dashboard (`report_dashboard`).\n\n## Non-Negotiable Guardrails\n- **Clean Context Window Isolation per Specialist Agent (Cross-IDE)**: Never execute two distinct specialist roles (e.g., `agent-qa-engineer` and `agent-developer`, or `agent-developer` and `agent-adversary`) in the same conversation context window. In subagent-capable IDEs (Antigravity `invoke_subagent`, Claude Code `Task` / `.claude/agents/`, GitHub Copilot `@agent-*`), always spawn each specialist in a fresh isolated subagent context that reads state strictly from disk (`specs/`, `handoff.yaml`, `get_active_handoff`). In chat-based IDEs (Cursor, Windsurf, Aider), persist all artifacts to disk, emit the Workflow Handoff block, and instruct the user to open a **New Chat** (or `/clear`) before running the next specialist.\n- **2-Checkpoint Quality & Adversarial Protocol**: Enforce transversal quality and red-team checks exclusively at **Checkpoint 1 (Specification & Design)** and **Checkpoint 2 (Implementation & Pre-PR)**. Enforce the **Anti-Ping-Pong Guardrail**: maximum 2 correction rounds per checkpoint; if unresolved blocking defects persist, escalate to the Human Tech Lead.\n- **Single-Use-Case Scope (`1 UC-* = 1 CHG-*`)**: Each SDD change (`CHG-*`) must deliver at most 1 Use Case (`UC-*`). Bundling multiple `UC-*` into one `handoff.yaml` is blocked by `verify_governance` (`Greenfield monolith detected`). Scale throughput via parallel `CHG-*` branches/worktrees, not monolithic changes.\n- **100% Design & Handoff Coverage in `tasks.md`**: Every `CMP-*`, `FR-*`, `SEC-REQ-*`, and `PRIV-REQ-*` declared in `design.md` and `handoff.yaml` must be explicitly assigned to at least one `TSK-*` in `tasks.md`.'
19760
+ body: '## Purpose & Trigger\nUse this skill for **every evolutionary increment (new feature, enhancement, refactoring, or bugfix)** on a project that already has an AI-SDLC baseline. It enforces strict 360\xB0 traceability from upstream ProductShape (`UC-*`, `FR-*`) through SDD (`CHG-*`, `HOF-*`, `TSK-*`), TDD implementation, the 12 Quality Gates, and baseline promotion under the **streamlined 2-checkpoint architecture**.\n\n## End-to-End Execution Playbook\n\n### Step 0: Branch-Before-Change, Pre-Change Draft PR & Scope Triage (`patch | standard | critical`)\n1. **Branch-Before-Change & Pre-Change Draft PR/MR**: Verify the active branch (`git branch --show-current`). If currently on a protected branch (`main`, `master`, `release/*`), create and switch to a dedicated working branch (`feat/<CHG-ID>-<slug>` or `fix/<CHG-ID>-<slug>`) **before** modifying or creating any files. Whenever a remote repository is configured (`git remote`), check if an open PR/MR already exists for the branch (`gh pr view` / `gh pr list --head <branch>` on GitHub; `glab mr view` / `glab mr list --source-branch <branch>` on GitLab); if none exists, create an empty bootstrap commit (`git commit --allow-empty -m "chore(<CHG-ID>): open draft PR"`), push (`git push -u origin <branch>`), and open a **Draft PR/MR** (`gh pr create --draft` on GitHub; `glab mr create --draft` or `git push -o merge_request.create -o merge_request.draft` on GitLab) before making any changes (if no remote exists, work locally).\n2. Before editing any files or running `sdd_change_new`, inspect if the user explicitly specified the friction profile (`patch`, `standard`, or `critical`). If not, analyze the requested scope and ask the user via interactive question (`ask_question`) with the recommended option first (`(Recommended)`):\n - **`patch`**: low-risk isolated bugfixes ($\\le 2$ source files, zero Anti-Bypass paths), typos, or linter fixes on `patch/*` or `fix/*` branches (`npx aisdlc change new "<Title>" --id <chg-id> --profile patch` generates `spec.md` with `profile: patch` and `verification` block only, omitting `design.md`, `tasks.md`, and `handoff.yaml`; fast-track directly to Step 3 -> Step 4 -> Step 5 -> Step 6).\n - **`standard`**: functional increments, new/updated `UC-*` or `FR-*`, or contract changes (`proposal.md`, `spec.md`, `design.md`, `tasks.md`, and `handoff.yaml`).\n - **`critical`**: changes touching security enclaves (`SEC-ENC-*`), sensitive PII (`SPECIAL_CATEGORY` / `DPIA-*`), schemas, or migrations (full `standard` artifacts + `ADR-*` + mandatory STRIDE modeling and dual human approval).\n2. **Anti-Bypass Rule**: Never recommend or allow `patch` when the change touches protected paths (`schemas/`, `specs/security/`, `specs/privacy/`, `packages/security/`, `SEC-ENC-*`, `09_decisions/`, `migrations/`, or `*-policy.yaml`).\n\n### Step 1: Upstream Product, Security & Privacy Delta (`agent-product-analyst`, `agent-threat-modeler` & `agent-privacy-officer`)\n1. Inspect active changes via MCP tool `get_active_handoff` and verify whether the target Use Case (`UC-*`) or Requirement (`FR-*`) already exists in `specs/product/`.\n2. If the increment introduces new behavior or modifies business rules:\n - Use [`agent-product-analyst`](../agent-product-analyst/SKILL.md) (running a `/grill-me` Socratic Decision-Tree interview of 1\u20132 focused questions per turn with `(Recommended)` options and Read-First codebase exploration whenever design branches are open) to create or update the `UC-*`, `BR-*`, and `FR-*` (with `@nominal`, `@boundary`, `@invalid` Gherkin scenarios).\n - Run MCP tool `verify_duplicates` (`npx aisdlc verify duplicates`) to block lexical or functional collisions early.\n - Use [`agent-threat-modeler`](../agent-threat-modeler/SKILL.md) to model any new `ABUSE-*` and `SEC-REQ-*` mitigations.\n - Hand off to [`agent-privacy-officer`](../agent-privacy-officer/SKILL.md) to model any personal data assets (`DATA-ASSET-*`), GDPR Art. 30 processing activities (`PRIV-ROPA-*`), and privacy requirements (`PRIV-REQ-*`).\n3. Run `sdd_check_fix` (`npx aisdlc check --fix`) to sync Gherkin `.feature` files and validate schemas.\n\n### Step 2: SDD Change Scaffolding & Architectural Design (`agent-system-architect`)\n1. Scaffold the bounded change package and its cryptographic `handoff.yaml` sidecar (`HOF-*`) targeting **a single Use Case (`1 UC-* = 1 CHG-*`)** using MCP tool `sdd_change_new` (or CLI `npx aisdlc change new "<Title>" --from <UC-ID> --profile <patch|standard|critical>`).\n2. Use [`agent-system-architect`](../agent-system-architect/SKILL.md) to complete `specs/changes/active/<chg-id>/design.md` and `specs/architecture/CMP-*.md`:\n - Structure `CMP-*` and `design.md` around **Modular & Hexagonal Architecture (Ports & Adapters)** (Domain/Application Core + Inbound/Outbound Ports) and define **Contract-First** shared interfaces/DTOs/schemas when spanning Frontend and Backend.\n - Map 100% of `handoff.yaml` requirements and components to atomic tasks (`TSK-*`) in `tasks.md` with explicit `autonomy-mode` (`AUTONOMOUS`, `HUMAN_REVIEW_PLAN`, `HIGH_RISK_MANUAL`, `AMBIGUOUS`).\n3. Run `verify_governance` and `verify_friction` to validate task governance, 100% design/handoff scope coverage, and anti-monolith limits (`maxUseCasesPerChange = 1`).\n4. Commit the SDD scaffolding and specification/design artifacts on the working branch and push (`git push`) to the open Draft PR.\n\n### Checkpoint 1: Specification & Design Checkpoint (`agent-auditor` $\\parallel$ `agent-adversary`)\n1. Concurrently invoke `agent-auditor` (evaluating YAGNI, interface bloat, and task completeness) and `agent-adversary` (probing contradictions, race conditions, and architecture SPOFs) in clean, isolated context windows over the combined upstream deliverables.\n2. Consolidate findings and remediate any blocking defects (max 2 correction rounds), committing and pushing updates (`git push`) to the open Draft PR.\n3. **Human Approval Gate 1**: Present the specification and architectural design (reviewable directly in the Draft PR) for Human Tech Lead / PO sign-off before proceeding to code.\n\n### Step 3: Streamlined Git Branch Navigation & Multi-Agent Parallelism\n1. Confirm you are on the working branch via `npx aisdlc git checkout <TSK-ID>` (`pnpm run git:checkout <TSK-ID>`).\n - **Default Streamlined Mode**: Ensures the agent is on `feat/<CHG-ID>-<slug>` (already created in Step 0).\n - Granular task traceability is maintained via commit trailers (`Task-ID: TSK-*`, `Parent-Ref: CHG-*`) and commit scopes (`feat(TSK-001): ...`), enabling per-task rework KPIs without branch explosion.\n2. **Parallel Agent Execution Rules (`--parallel`)**:\n - When concurrent multi-agent implementation is configured (e.g. Backend adapter $\\parallel$ Frontend UI), `aisdlc git checkout <TSK-ID> --parallel` creates isolated worktrees (`task/<PARENT-ID>/<TSK-ID>-<slug>`).\n\n### Step 4: Red Phase \u2014 Failing Test Suites (`agent-qa-engineer`)\n1. Delegate to or adopt [`agent-qa-engineer`](../agent-qa-engineer/SKILL.md).\n2. Run `npx aisdlc gherkin extract` (`sdd_check_fix`) to deterministically extract `.feature` files from Markdown specifications, and run `npx aisdlc gherkin scaffold <feature-file> --lang <lang> --fallback` to scaffold RED step definitions in `tests/steps/` (never hand-author `.feature` or step skeletons).\n3. Author FAILING (red) assertions inside the scaffolded step definitions, plus integration/contract tests and mandatory E2E/UI tests (evaluating Happy Path, Sad Path, and Boundary Values for every interactive UI element, plus `UC-*` alternative/exception flows) in `tests/`.\n4. Commit failing test suites with trailer `Task-ID: TSK-YYY` and push (`git push`) to the Draft PR (CI workflows skip heavy checks while the PR is in `draft` state).\n5. Run `npx aisdlc verify gherkin` and `verify_testing` (`npx aisdlc verify testing`) to confirm deterministic Gherkin/step sync and requirement test coverage before writing production code.\n\n### Step 5: Green & Refactor Phase \u2014 Surgical Implementation (`agent-developer` / `agent-devops`)\n1. Delegate application tasks (`src/`, `tests/`) to [`agent-developer`](../agent-developer/SKILL.md). For tasks (`TSK-*`) targeting CI/CD workflows (`.github/workflows/`), `Dockerfile*`, IaC manifests, or support scripts (`scripts/`), delegate to [`agent-devops`](../agent-devops/SKILL.md) under its strict non-invasion guardrail on `src/`.\n2. Execute the **Mandatory 4-Step Implementation Order**:\n - (1) Modular & Hexagonal Architecture (Ports & Adapters) boundaries, (2) Inter-module interfaces/ports first, (3) Concrete component implementation (zero fake-wiring; all 4 UI states `Loading`, `Empty`, `Error`, `Nominal`), (4) Exhaustive per-function unit tests in `tests/unit/` (Happy Path, Sad Path, Boundary, Null/Empty).\n3. Commit implementations with trailer `Task-ID: TSK-YYY` and push (`git push`) incrementally to the Draft PR.\n4. Verify any new dependency against `license-policy.yaml` via `verify_licenses` and PII handling against `privacy-policy.yaml` via `verify_privacy`.\n5. Run `sdd_check_fix` (`npx aisdlc check --fix`) and `verify_quality` (`npx aisdlc verify quality`) until Cyclomatic Complexity $\\le 10$, Cognitive $\\le 15$, Maintainability $\\ge 50$, and Function Length $\\le 40$ lines pass with zero violations.\n6. Run `verify` (`npx aisdlc verify all`) confirming all **12 Quality Gates** pass green.\n\n### Checkpoint 2: Implementation & Pre-PR Checkpoint (`agent-auditor` $\\parallel$ `agent-adversary`)\n1. **Functional Validation (`agent-expert-user`)**: Validate live runtime behavior against `UC-*` and `FR-*` criteria (including `0` browser `console.error`/`pageerror` on UI changes).\n2. Concurrently invoke `agent-auditor` (verifying Clean Code, SOLID, DRY, AST thresholds, and eliminating unnecessary abstractions) and `agent-adversary` (performing conceptual test mutation to verify QA tests catch defects, and attacking boundary inputs with concrete PoC breaks).\n3. Consolidate findings and remediate in `agent-developer` (max 2 correction rounds), committing and pushing (`git push`) fixes to the Draft PR.\n4. Run `npx aisdlc verify all` to confirm final green status, then run `report_quality` (`npx aisdlc report quality`) and `report_dashboard` (`npx aisdlc report dashboard`) to compute local reports and the interactive dashboard for review (**do NOT stage or commit global `reports/*` files on `feat/CHG-*` branches** to prevent merge conflicts across parallel PRs).\n\n### Step 6: Human PR Sign-Off, Baseline Integration & Release Closure\n1. **Ready for Review & Feature Merge (`feat/CHG-*` \u2794 `main` or `release/vX.Y.Z`)**:\n - Update the Draft PR body with the Plan Execution Matrix (`[X]` vs `[O]`), KPI summary (`kpi_pr` including per-task rework metrics), and local verification/report summary (`reports/QUALITY_REPORT.md`, `reports/TRACEABILITY_MATRIX.md`, `reports/dashboard.html` inspected locally, without committing global `reports/*` to the feature PR), then run `gh pr ready` (GitHub) or `glab mr update --ready` (GitLab) to mark the PR/MR **Ready for Review** (triggering full CI verification) and present it to the human Tech Lead for review and approval (never auto-approve or auto-merge).\n2. **Canonical Baseline Promotion**: Once tasks in `tasks.md` are marked `COMPLETED` and approved, promote the change into the canonical baseline via MCP tool `sdd_integrate` (`npx aisdlc sdd integrate --auto`), which synchronizes the normative text and BDD Gherkin scenarios into `FR-*` / `UC-*` / `SEC-REQ-*` / `PRIV-REQ-*`, bumps minor SemVer when updating active specs, resyncs PDaC SHA-256 digests, promotes `FR-*` to `active`, links `CMP-*`, and archives the change under `specs/changes/completed/` (or `npx aisdlc sdd integrate --all` to consolidate all completed changes).\n3. **Semantic Versioning Promotion**:\n - **Profile-Driven SemVer Mapping**: Bump the version of the software under development according to the increment\'s friction profile:\n - `patch` -> bump **Patch** digit (`x.y.Z`)\n - `standard` -> bump **Minor** digit (`x.Y.0`)\n - `critical` -> bump **Major** digit (`X.0.0`)\n - **Highest-Precedence Consolidation (Multiple Pending Increments)**: When multiple pending increments exist in the active branch or release batch being integrated, apply **ONLY a single version bump corresponding to the highest rank** (`critical` / Major > `standard` / Minor > `patch` / Patch), ignoring unrelated historical directories.\n - Tag the merge with the computed SemVer bump (`patch`, `minor`, `major`) and refresh the interactive dashboard (`report_dashboard`).\n\n## Non-Negotiable Guardrails\n- **Clean Context Window Isolation per Specialist Agent (Cross-IDE)**: Never execute two distinct specialist roles (e.g., `agent-qa-engineer` and `agent-developer`, or `agent-developer` and `agent-adversary`) in the same conversation context window. In subagent-capable IDEs (Antigravity `invoke_subagent`, Claude Code `Task` / `.claude/agents/`, GitHub Copilot `@agent-*`), always spawn each specialist in a fresh isolated subagent context that reads state strictly from disk (`specs/`, `handoff.yaml`, `get_active_handoff`). In chat-based IDEs (Cursor, Windsurf, Aider), persist all artifacts to disk, emit the Workflow Handoff block, and instruct the user to open a **New Chat** (or `/clear`) before running the next specialist.\n- **2-Checkpoint Quality & Adversarial Protocol**: Enforce transversal quality and red-team checks exclusively at **Checkpoint 1 (Specification & Design)** and **Checkpoint 2 (Implementation & Pre-PR)**. Enforce the **Anti-Ping-Pong Guardrail**: maximum 2 correction rounds per checkpoint; if unresolved blocking defects persist, escalate to the Human Tech Lead.\n- **Single-Use-Case Scope (`1 UC-* = 1 CHG-*`)**: Each SDD change (`CHG-*`) must deliver at most 1 Use Case (`UC-*`). Bundling multiple `UC-*` into one `handoff.yaml` is blocked by `verify_governance` (`Greenfield monolith detected`). Scale throughput via parallel `CHG-*` branches/worktrees, not monolithic changes.\n- **100% Design & Handoff Coverage in `tasks.md`**: Every `CMP-*`, `FR-*`, `SEC-REQ-*`, and `PRIV-REQ-*` declared in `design.md` and `handoff.yaml` must be explicitly assigned to at least one `TSK-*` in `tasks.md`.'
19732
19761
  },
19733
19762
  {
19734
19763
  id: "aisdlc-inspector",
@@ -19736,7 +19765,7 @@ var LIFECYCLE_SKILL_TEMPLATES = [
19736
19765
  description: "Interactive technical inspector and graph navigator across AI-SDLC requirements (FR-*, QR-*, SEC-REQ-*, PRIV-REQ-*), architecture components (CMP-*, ADR-*), SDD tasks (TSK-*, CHG-*), test suites, and source code files. Provides concise technical summaries, upstream/downstream traceability mappings, and selectable navigation loops.",
19737
19766
  references: "[`process/06_spec_driven_development.md`](../../process/06_spec_driven_development.md), [`process/08_citation_contract_and_drift.md`](../../process/08_citation_contract_and_drift.md), and [`process/10_quality_management_and_release_gates.md`](../../process/10_quality_management_and_release_gates.md).",
19738
19767
  globs: "specs/**, architecture/**, src/**, packages/**, tests/**",
19739
- body: '## Purpose & Trigger\nUse this skill whenever a user or agent requests to **inspect, explain, trace, or navigate** any element of the system:\n- **Requirements & Product:** `UC-*`, `FR-*`, `QR-*`, `BR-*`, `BC-*`, `ABUSE-*`, `DATA-ASSET-*`, `PRIV-ROPA-*`, `ARCH-INTRO-*`.\n- **Architecture & Decisions:** `CMP-*`, `ADR-*`, `SEC-ENC-*`.\n- **SDD & Tasks:** `TSK-*`, `CHG-*`, `HOF-*`.\n- **Source Code Files:** `src/**/*.ts`, `packages/**/*.ts`, etc.\n- **Verification Suites:** `.feature` BDD files, unit/integration specs.\n\nThis skill provides **impact analysis, context comprehension, and interactive graph exploration** with zero guesswork.\n\n---\n\n## Interactive Execution Playbook\n\n### Step 1: Target Resolution\n1. Extract the target identifier or file path indicated by the user (e.g., `FR-001`, `CMP-CORE`, `src/auth/jwt.ts`, `TSK-001`).\n2. If no target was specified, prompt the user with common anchor options (e.g., primary Use Cases `UC-*`, Core Components `CMP-*`, or an open input).\n\n### Step 2: Deterministic Inspection Engine\nExecute the inspection via MCP or CLI:\n* **Primary (MCP Tool):**\n Invoke MCP tool `inspect_element`:\n ```json\n {\n "query": "<target-id-or-path>"\n }\n ```\n* **Fallback (CLI):**\n Run command:\n ```bash\n npx aisdlc inspect "<target-id-or-path>" --json\n ```\n\n---\n\n### Step 3: Complete Technical Report Delivery (MANDATORY FIRST)\n> [!IMPORTANT]\n> **CRITICAL UX DIRECTIVE (Report First, Question After):**\n> The agent MUST always render and deliver the complete, detailed technical report to the user in visible text **BEFORE** prompting for the next element. Never invoke a blocking dialog or selector on an empty turn without the user having the full inspection visible.\n\nThe rendered response MUST contain:\n1. **Header:** Artifact ID, Title, Type (`requirement`, `component`, `source-file`, `task`), Layer (`product`, `architecture`, `code`, `test`, `governance`), and Status (`COMPLIANT`, `ORPHAN`, `DRIFT`).\n2. **File Location:** Clickable markdown link to the file.\n3. **Technical Summary:** Concise explanation (2\u20134 sentences) of what the element does, its responsibilities, key interfaces, and consolidated evolutionary increments (`### Incremento (CHG-*)`). If a canonical requirement still shows boilerplate text despite completed changes in `specs/changes/`, run `sdd_integrate` with `all: true` (`npx aisdlc sdd integrate --all`) first so canonical requirements reflect the consolidated specification.\n4. **Upstream Relations (Derives from / Mitigates / Governed by):**\n - \u2197\uFE0F List all parent requirements, use cases, business rules, or decisions that justify this element.\n5. **Downstream Relations (Satisfied by / Implemented in / Verified by):**\n - \u2198\uFE0F List all components implementing it, tasks developing it, and test suites validating it.\n6. **Numbered Navigable Items:** Clear, numbered list of connected nodes with their ID, type, and relationship.\n\n---\n\n### Step 4: Subsequent Navigation Formulation & Selector\nOnce the technical report is fully constructed:\n1. Formulate the closing question inviting the user to continue navigating the graph.\n2. If using the interactive selector (`ask_question`), execute it **in conjunction with or immediately following the complete report text**:\n * **Question:** *"\xBFHacia qu\xE9 elemento conectado deseas continuar la exploraci\xF3n?"*\n * **Options:** Formatted list of related elements from `relatedItems`:\n - `(Recommended) <ID-1> (<Type-1> \u2014 <relation-1>)`\n - `<ID-2> (<Type-2> \u2014 <relation-2>)`\n - `<ID-N> (<Type-N> \u2014 <relation-N>)`\n - `Terminar exploraci\xF3n`\n3. **Navigation Loop:**\n - When the user selects or indicates an element (either via UI selector or plain text/number), repeat **Step 2** with the new selection.\n - If the user selects `Terminar exploraci\xF3n`, summarize the traversal journey and conclude.'
19768
+ body: '## Purpose & Trigger\nUse this skill whenever a user or agent requests to **inspect, explain, trace, or navigate** any element of the system:\n- **Requirements & Product:** `UC-*`, `FR-*`, `QR-*`, `BR-*`, `BC-*`, `ABUSE-*`, `DATA-ASSET-*`, `PRIV-ROPA-*`, `ARCH-INTRO-*`.\n- **Architecture & Decisions:** `CMP-*`, `ADR-*`, `SEC-ENC-*`.\n- **SDD & Tasks:** `TSK-*`, `CHG-*`, `HOF-*`.\n- **Source Code Files:** `src/**/*.ts`, `packages/**/*.ts`, etc.\n- **Verification Suites:** `.feature` BDD files, unit/integration specs.\n\nThis skill provides **impact analysis, context comprehension, and interactive graph exploration** with zero guesswork.\n\n---\n\n## Interactive Execution Playbook\n\n### Step 1: Target Resolution\n1. Extract the target identifier or file path indicated by the user (e.g., `FR-001`, `CMP-CORE`, `src/auth/jwt.ts`, `TSK-001`).\n2. If no target was specified, prompt the user with common anchor options (e.g., primary Use Cases `UC-*`, Core Components `CMP-*`, or an open input).\n\n### Step 2: Deterministic Inspection Engine\nExecute the inspection via MCP or CLI:\n* **Primary (MCP Tool):**\n Invoke MCP tool `inspect_element`:\n ```json\n {\n "query": "<target-id-or-path>"\n }\n ```\n* **Fallback (CLI):**\n Run command:\n ```bash\n npx aisdlc inspect "<target-id-or-path>" --json\n ```\n\n---\n\n### Step 3: Complete Technical Report Delivery (MANDATORY FIRST)\n> [!IMPORTANT]\n> **CRITICAL UX DIRECTIVE (Report First, Question After):**\n> The agent MUST always render and deliver the complete, detailed technical report to the user in visible text **BEFORE** prompting for the next element. Never invoke a blocking dialog or selector on an empty turn without the user having the full inspection visible.\n\nThe rendered response MUST contain:\n1. **Header:** Artifact ID, Title, Type (`requirement`, `component`, `source-file`, `task`), Layer (`product`, `architecture`, `code`, `test`, `governance`), and Status (`COMPLIANT`, `ORPHAN`, `DRIFT`).\n2. **File Location:** Clickable markdown link to the file.\n3. **Technical Summary:** Concise explanation (2\u20134 sentences) of what the element does, its responsibilities, key interfaces, and consolidated evolutionary increments (`### Incremento (CHG-*)`). If a canonical requirement still shows boilerplate text despite completed changes in `specs/changes/`, run `sdd_integrate` with `all: true` (`npx aisdlc sdd integrate --all`) first so canonical requirements reflect the consolidated specification.\n4. **Upstream Relations (Derives from / Mitigates / Governed by):**\n - \u2197\uFE0F List all parent requirements, use cases, business rules, or decisions that justify this element.\n5. **Downstream Relations (Satisfied by / Implemented in / Verified by):**\n - \u2198\uFE0F List all components implementing it, tasks developing it, and test suites validating it.\n6. **Numbered Navigable Items:** Clear, numbered list of connected nodes with their ID, type, and relationship.\n\n---\n\n### Step 4: Subsequent Navigation Formulation & Selector\nOnce the technical report is fully constructed:\n1. Formulate the closing question inviting the user to continue navigating the graph.\n2. If using the interactive selector (`ask_question`), execute it **in conjunction with or immediately following the complete report text**:\n * **Question:** *"\xBFHacia qu\xE9 elemento conectado deseas continuar la exploraci\xF3n?"*\n * **Options:** Formatted list of related elements from `relatedItems`:\n - `(Recommended) <ID-1> (<Type-1> \u2014 <relation-1>)`\n - `<ID-2> (<Type-2> \u2014 <relation-2>)`\n - `<ID-N> (<Type-N> \u2014 <relation-N>)`\n - `Auditar este elemento (<ID>) con agent-auditor + agent-adversary (Handoff \u2192 aisdlc-spot-audit)`\n - `Terminar exploraci\xF3n`\n3. **Navigation & Audit Handoff Loop:**\n - When the user selects or indicates an element (either via UI selector or plain text/number), repeat **Step 2** with the new selection.\n - If the user selects `Auditar este elemento (<ID>) con agent-auditor + agent-adversary (Handoff \u2192 aisdlc-spot-audit)`, emit a Workflow Handoff transitioning directly into [`aisdlc-spot-audit`](../aisdlc-spot-audit/SKILL.md) targeting `<ID>`.\n - If the user selects `Terminar exploraci\xF3n`, summarize the traversal journey and conclude.'
19740
19769
  },
19741
19770
  {
19742
19771
  id: "aisdlc-doctor-migrate",
@@ -19769,6 +19798,14 @@ var LIFECYCLE_SKILL_TEMPLATES = [
19769
19798
  references: "[`process/06_spec_driven_development.md`](../../process/06_spec_driven_development.md) and [`process/10_quality_management_and_release_gates.md`](../../process/10_quality_management_and_release_gates.md).",
19770
19799
  globs: "tests/features/**, tests/steps/**, specs/**",
19771
19800
  body: "## Purpose & Trigger\nUse this skill whenever **authoring BDD acceptance criteria, extracting `.feature` files, or scaffolding RED step definitions** in TypeScript, Python, Go, Rust, or C#.\n\n## End-to-End Execution Playbook\n\n### Step 1: Author Gherkin Inside Markdown Specs\n1. Never create or edit `tests/features/*.feature` files manually.\n2. Author ```gherkin blocks inside `FR-*`, `SEC-REQ-*`, `PRIV-REQ-*`, or `specs/changes/active/<chg-id>/spec.md` with tags `@<ID> @automated @regression` and `@nominal`, `@boundary`, `@invalid` scenarios.\n\n### Step 2: Deterministic Extraction & Polyglot Step Scaffolding\n1. Run `npx aisdlc gherkin extract --all` (or `npx aisdlc check --fix`) to generate `tests/features/*.feature` deterministically.\n2. Scaffold failing RED step definitions in `tests/steps/` via:\n `npx aisdlc gherkin scaffold <feature-file> --lang <ts|py|go|rs|cs> --fallback`\n (omit `--fallback` when the native BDD generator is installed).\n3. Implement real failing assertions inside the scaffolded step files.\n\n### Step 3: Synchronization Gate\n1. Run `npx aisdlc verify gherkin` and `npx aisdlc verify testing` to confirm 100% Markdown-to-`.feature` parity and scenario coverage."
19801
+ },
19802
+ {
19803
+ id: "aisdlc-spot-audit",
19804
+ title: "360\xB0 Polymorphic Element Auditor & Adversarial Stress Playbook",
19805
+ description: "Audit any system element on demand \u2014 product specs (UC-*, FR-*, QR-*, BR-*), security/privacy artifacts (ABUSE-*, SEC-REQ-*, DATA-ASSET-*, PRIV-REQ-*), architecture components (CMP-*, ADR-*), SDD tasks (CHG-*, TSK-*), test suites, or source code files \u2014 by combining deterministic graph inspection (inspect_element) with parallel isolated evaluations by agent-auditor and agent-adversary and emitting a canonical Workflow Handoff.",
19806
+ references: "[`process/01_governance_and_roles.md`](../../process/01_governance_and_roles.md), [`process/09_agent_protocols.md`](../../process/09_agent_protocols.md), [`process/10_quality_management_and_release_gates.md`](../../process/10_quality_management_and_release_gates.md), and [`templates/workflow/agent-handoff.template.md`](../../templates/workflow/agent-handoff.template.md).",
19807
+ globs: "specs/**, architecture/**, security/**, privacy/**, src/**, packages/**, tests/**",
19808
+ body: '## Purpose & Trigger\nUse this skill whenever a user or upstream workflow (including a handoff from [`aisdlc-inspector`](../aisdlc-inspector/SKILL.md)) requests to **audit, stress-test, red-team, or challenge any single element or set of elements** on demand without running a full SDD lifecycle:\n- **Product & Vision:** `ARCH-INTRO-*`, `ACT-*`, `JRN-*`, `UC-*`, `FR-*`, `QR-*`, `BR-*`.\n- **Cybersecurity & Privacy:** `ACT-THREAT-*`, `ABUSE-*`, `SEC-REQ-*`, `DATA-ASSET-*`, `PRIV-ROPA-*`, `PRIV-REQ-*`.\n- **Architecture & Decisions:** `CMP-*`, `ADR-*`, `ARCH-DEP-*`, `SEC-ENC-*`.\n- **SDD Changes & Tasks:** `CHG-*`, `HOF-*`, `TSK-*`.\n- **Source Code & Infrastructure:** `src/**/*`, `packages/**/*`, `.github/workflows/*.yml`, `Dockerfile*`.\n- **Verification Suites:** `tests/**/*.spec.*`, `tests/features/*.feature`, `tests/steps/*`.\n\n---\n\n## End-to-End Execution Playbook\n\n### Step 1: Target Resolution & 360\xB0 Context Hydration (`inspect_element`)\n1. Extract the target identifier or file path indicated by the user (e.g., `UC-001`, `FR-002`, `CMP-CORE`, `src/auth/jwt.ts`, `tests/unit/auth.spec.ts`). If none was provided, prompt via `ask_question` with primary Use Cases (`UC-*`), Core Components (`CMP-*`), or active SDD changes (`CHG-*`).\n2. Run MCP tool `inspect_element` (`{"query": "<target-id-or-path>"}`) or CLI `npx aisdlc inspect "<target-id-or-path>" --json` to hydrate:\n - Canonical ID, layer (`product`, `architecture`, `security`, `privacy`, `code`, `test`, `governance`), file path, and Unix LF SHA-256 citation digest.\n - Traceability status (`COMPLIANT`, `ORPHAN`, `DRIFT`) and immediate upstream (\u2197\uFE0F) and downstream (\u2198\uFE0F) neighbors.\n3. Read the target file and its directly connected artifacts from disk.\n\n### Step 2: Layer-Specific Deterministic Pre-Audit Gates\nExecute deterministic verifiers matching the element\'s layer before invoking specialist subagents:\n- **Product / Security / Privacy / Architecture Specs:** Run `verify_schemas`, `verify_pdac`, and `verify_duplicates` (plus `verify_privacy` for `DATA-ASSET-*` / `PRIV-*`).\n- **Source Code (`src/**`, `packages/**`):** Run `verify_quality` (`quality-policy.yaml` AST metrics), `verify_security` (secrets & SAST), `verify_privacy` (PII leaks), and `verify_licenses`.\n- **Verification Suites (`tests/**`, `*.feature`):** Run `npx aisdlc verify gherkin` and `verify_testing`.\n- **SDD Changes & Tasks (`CHG-*`, `TSK-*`):** Run `verify_governance`, `verify_friction`, and `verify_traceability`.\n\n### Step 3: Concurrent Polymorphic Subagent Evaluation (`agent-auditor` $\\parallel$ `agent-adversary`)\n> [!IMPORTANT]\n> **MANDATORY CLEAN CONTEXT WINDOW ISOLATION (Commandment #6):**\n> Never role-play `agent-auditor` and `agent-adversary` inside the same context window. In subagent-capable IDEs (Antigravity `invoke_subagent`, Claude Code `Task`, Copilot `@agent-*`), spawn two isolated subagents in parallel passing the target file path, SHA-256 digest, and upstream/downstream context from Steps 1\u20132.\n\n1. **Branch A \u2014 [`agent-auditor`](../agent-auditor/SKILL.md) (Constructive Quality, YAGNI & Compliance Audit):**\n - **Product (`UC-*`, `FR-*`, `BR-*`):** Check for ambiguity, missing `@nominal`/`@boundary`/`@invalid` Gherkin scenarios, and enforce End-to-End MVP Completeness (rejecting library-only MVPs when human actors `ACT-*` exist, verifying UI modality and master-data CRUD prerequisites).\n - **Architecture (`CMP-*`, `ADR-*`):** Enforce Hexagonal Architecture (Ports & Adapters), YAGNI, Contract-First shared schemas, and persistent `CMP-OBSERVABILITY` (`OBS_STORAGE_PATH`, `flush()`, `ITelemetryExporter`).\n - **Source Code (`src/**`):** Enforce `quality-policy.yaml` thresholds ($CC \\le 10$, $Cognitive \\le 15$, $MI \\ge 50$, $Lines \\le 40$), SOLID/DRY, 1:1 method unit test (`tests/unit/`) and UI component test (`tests/e2e/` or `tests/ui/`) coverage, zero fake-wiring, all 4 UI states (`Loading`, `Empty`, `Error`, `Nominal`), modern visual aesthetics (Bootstrap 5 / modern Desktop / rich ANSI Console), and SPDX license compliance.\n - **Tests (`tests/**`):** Verify deterministic Gherkin extraction/scaffolding and per-interactive-element Happy/Sad/Boundary coverage.\n - **Classification:** Label every finding as `[BLOCKING_DEFECT]`, `[SIMPLIFICATION]`, or `[ADVISORY]`.\n\n2. **Branch B \u2014 [`agent-adversary`](../agent-adversary/SKILL.md) (Red Team, Chaos & Proof-of-Break Audit):**\n - **Product (`UC-*`, `FR-*`, `BR-*`):** Uncover contradictory business rules, unreachable states, logic traps, and unmodeled abuse paths.\n - **Architecture (`CMP-*`, `ADR-*`):** Simulate race conditions, cascading failures, resource exhaustion, Single Points of Failure (SPOFs), and enclave boundary bypasses.\n - **Tests (`tests/**`):** Perform **Conceptual Mutation Testing** \u2014 provide concrete mutant production code showing how a broken or empty implementation would falsely pass the test suite green.\n - **Source Code & Security (`src/**`, `SEC-REQ-*`):** Attack boundary inputs, concurrency races, memory/type edge breaks, injection vectors (SQLi, XSS, command, path traversal, prompt injection), secret/PII leaks, and auth bypasses with **CVSS v3.1** scoring.\n - **Mandatory 4-Part Proof-of-Break:** Every adversarial finding MUST include: *(1) Target Entity ID & SHA-256*, *(2) Failure Scenario / Attack Vector*, *(3) Concrete Reproducible PoC (input payload, sequence breakdown, or mutant code)*, and *(4) Expected Impact & CVSS v3.1 score*.\n\n### Step 4: Consolidated Audit Report & Canonical Workflow Handoff\n> [!IMPORTANT]\n> **CRITICAL UX DIRECTIVE (Report First, Handoff & Selector After):**\n> Always render the complete Consolidated Audit Report in visible text **BEFORE** presenting the interactive follow-up selector.\n\n1. **Render the Consolidated Audit Report:**\n - **1. Audited Target Header:** Artifact ID / Path, Layer, Traceability Status (`COMPLIANT` / `ORPHAN` / `DRIFT`), and Unix LF SHA-256 digest.\n - **2. Deterministic Gate Summary:** Pass/Fail status of Step 2 verifiers.\n - **3. `agent-auditor` Findings Matrix:** Grouped by `[BLOCKING_DEFECT]`, `[SIMPLIFICATION]`, and `[ADVISORY]`.\n - **4. `agent-adversary` Proof-of-Break Matrix:** Findings with Attack Vector, Concrete PoC, and CVSS v3.1 severity.\n - **5. Overall Audit Verdict:** `\u2705 PASS` (0 blocking defects, 0 Critical/High PoC breaks), `\u26A0\uFE0F PASS_WITH_ADVISORIES` (only simplifications/advisories), or `\u274C BLOCKED` ($ge 1$ `[BLOCKING_DEFECT]` or Critical/High PoC break).\n\n2. **Emit Canonical Workflow Handoff Block (`templates/workflow/agent-handoff.template.md`):**\n Emit the `### \u{1F504} Workflow Handoff` section routing remediation to the target element\'s responsible owner agent (bounded by the **Anti-Ping-Pong Guardrail**: maximum 1 remediation return round):\n - `UC-*` / `FR-*` / `QR-*` / `BR-*` / `JRN-*` $\\rightarrow$ `agent-product-analyst` (or `agent-expert-user`)\n - `ACT-THREAT-*` / `ABUSE-*` / `SEC-REQ-*` $\\rightarrow$ `agent-threat-modeler`\n - `DATA-ASSET-*` / `PRIV-ROPA-*` / `PRIV-REQ-*` $\\rightarrow$ `agent-privacy-officer`\n - `CMP-*` / `ADR-*` / `ARCH-DEP-*` $\\rightarrow$ `agent-system-architect`\n - `tests/**` / `*.feature` $\\rightarrow$ `agent-qa-engineer`\n - `src/**` / `packages/**` $\\rightarrow$ `agent-developer`\n - `.github/workflows/**` / `Dockerfile*` / `scripts/**` $\\rightarrow$ `agent-devops`\n\n3. **Interactive Handoff & Navigation Selector (`ask_question`):**\n Offer the user the next action via `ask_question`:\n - `(Recommended) Ejecutar Handoff de remediaci\xF3n con <owner-agent> (corregir hallazgos detectados)` (or `(Recommended) Volver a aisdlc-inspector para explorar nodos conectados` if verdict is `\u2705 PASS`)\n - `Abrir incremento gobernado SDD (Handoff \u2192 aisdlc-increment)`\n - `Auditar un elemento conectado (<related-ID-1>, <related-ID-2>, ...)`\n - `Inspeccionar grafo 360\xB0 del elemento (Handoff \u2192 aisdlc-inspector)`\n - `Finalizar auditor\xEDa`\n\n## Non-Negotiable Guardrails\n- **Strict Non-Invasion:** Neither `aisdlc-spot-audit`, `agent-auditor`, nor `agent-adversary` may directly edit or patch the audited files during the audit turn. All modifications happen after the human approves the remediation Handoff to the owner specialist agent.\n- **Zero Abstract Criticism:** Every adversarial finding without a concrete PoC and every quality finding without a specific line/section citation must be discarded before rendering the report.'
19772
19809
  }
19773
19810
  ];
19774
19811
  function formatSkillReferencesSection(skillId) {
@@ -19862,7 +19899,7 @@ This file defines the operational rules and specialized role mappings for Google
19862
19899
 
19863
19900
  | Role | Mission (summary) |
19864
19901
  | :--- | :--- |
19865
- | \`agent-product-analyst\` | Phase 0 NAF v4 Vision (\`ARCH-INTRO-*\`, up to 15-question discovery defining UI modality & CRUD prerequisites) + ProductShape (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); \`status\` starts \`draft\`. |
19902
+ | \`agent-product-analyst\` | Phase 0 NAF v4 Vision (\`ARCH-INTRO-*\`, \`/grill-me\` Socratic decision-tree discovery interview of up to 15 questions \u2014 1\u20132 per turn with \`(Recommended)\` options \u2014 defining UI modality & CRUD prerequisites) + ProductShape (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); \`status\` starts \`draft\`. |
19866
19903
  | \`agent-threat-modeler\` | STRIDE/OWASP threat triad (\`ACT-THREAT-*\`, \`ABUSE-*\`, \`SEC-REQ-*\`); hands off to \`agent-privacy-officer\`. |
19867
19904
  | \`agent-privacy-officer\` | LINDDUN & GDPR/LOPDGDD privacy triad (\`DATA-ASSET-*\`, \`PRIV-ROPA-*\`, \`PRIV-REQ-*\`) + Pre-Merge \`verify:privacy\`. |
19868
19905
  | \`agent-system-architect\` | Modular arc42 + NAF v4 architecture (\`CMP-*\`, \`ARCH-DEP-001\`, Mermaid diagrams, \`ADR-*\`). |
@@ -19936,7 +19973,7 @@ Before finalizing changes or proposing commits:
19936
19973
  ## 4. Specialized Roles Mapping
19937
19974
  | Role | Mission (summary) |
19938
19975
  | :--- | :--- |
19939
- | \`agent-product-analyst\` | Phase 0 NAF v4 Enterprise & Capability Vision (\`ARCH-INTRO-*\`, up to 15-question discovery interview defining UI modality & CRUD prerequisites) + ProductShape artifacts (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); status starts \`draft\`. |
19976
+ | \`agent-product-analyst\` | Phase 0 NAF v4 Enterprise & Capability Vision (\`ARCH-INTRO-*\`, \`/grill-me\` Socratic decision-tree discovery interview of up to 15 questions \u2014 1\u20132 per turn with \`(Recommended)\` options \u2014 defining UI modality & CRUD prerequisites) + ProductShape artifacts (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); status starts \`draft\`. |
19940
19977
  | \`agent-threat-modeler\` | STRIDE/OWASP threat triad (\`ACT-THREAT-*\`, \`ABUSE-*\`, \`SEC-REQ-*\`); hands off to \`agent-privacy-officer\`. |
19941
19978
  | \`agent-privacy-officer\` | LINDDUN & GDPR/LOPDGDD privacy triad (\`DATA-ASSET-*\`, \`PRIV-ROPA-*\`, \`PRIV-REQ-*\`) + Pre-Merge \`verify:privacy\`. |
19942
19979
  | \`agent-system-architect\` | Modular arc42 + NAF v4 architecture (\`CMP-*\`, Mermaid diagrams, \`ADR-*\`). |
@@ -19986,7 +20023,7 @@ Every generated or modified artifact must follow canonical naming and satisfy it
19986
20023
  - Privacy Requirements: \`PRIV-REQ-[SUFFIX]\` (\`schemas/privacy/privacy-req.schema.json\`)
19987
20024
 
19988
20025
  ## 2. "AI as Scribe" Protocol Rules
19989
- - **Phase 0 (New Product Inception)**: When starting a new product, conduct an iterative discovery interview of up to 15 questions (or until the product scope and target UI modality \u2014 Web, Desktop, Console/CLI, Mobile, or API/SDK \u2014 are completely clear), instantiate \`ARCH-INTRO-001\` (\`naf-perspective: "Enterprise & Capability"\`) in \`draft\`, and derive \`ACT-*\`, \`UC-*\`, \`BR-*\`, \`FR-*\`, and \`QR-*\` (including runnable UI/CLI requirements and master-entity CRUD prerequisites when human actors exist).
20026
+ - **Phase 0 & Incremental Discovery (\`/grill-me\` Socratic Decision-Tree Interview)**: When starting a new product or resolving open product decisions, conduct a \`/grill-me\` Socratic decision-tree discovery interview of up to 15 questions (1\u20132 focused questions per turn with \`(Recommended)\` options, Read-First codebase inspection, constructive pushback, and No-Premature-Drafting Gate until the product scope and target UI modality \u2014 Web, Desktop, Console/CLI, Mobile, or API/SDK \u2014 are completely clear), instantiate \`ARCH-INTRO-001\` (\`naf-perspective: "Enterprise & Capability"\`) in \`draft\`, and derive \`ACT-*\`, \`UC-*\`, \`BR-*\`, \`FR-*\`, and \`QR-*\` (including runnable UI/CLI requirements and master-entity CRUD prerequisites when human actors exist).
19990
20027
  - **Initial Status**: Every new artifact strictly starts with \`status: draft\`. Creating artifacts directly as \`active\` or \`approved\` is forbidden.
19991
20028
  - **Gherkin Acceptance Criteria**: In every \`FR-*\` requirement, include a \`\`\`gherkin block with tags \`@<ID> @automated @regression\` and at least one \`Scenario Outline\` with an \`Examples\` data table.
19992
20029
  - **Security & Privacy Mitigations**: Every \`SEC-REQ-*\` must include negative Gherkin tests (\`@security @mitigation\`) and every \`PRIV-REQ-*\` must include Gherkin privacy scenarios (\`@privacy @gdpr\`).
@@ -20073,7 +20110,7 @@ npx tsx packages/cli/src/index.ts sdd integrate --change <chg-id>
20073
20110
 
20074
20111
  | Role | Mission (summary) |
20075
20112
  | :--- | :--- |
20076
- | \`agent-product-analyst\` | Phase 0 NAF v4 Vision (\`ARCH-INTRO-*\`, up to 15-question discovery defining UI modality & CRUD prerequisites) + ProductShape (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); \`status\` starts \`draft\`. |
20113
+ | \`agent-product-analyst\` | Phase 0 NAF v4 Vision (\`ARCH-INTRO-*\`, \`/grill-me\` Socratic decision-tree discovery interview of up to 15 questions \u2014 1\u20132 per turn with \`(Recommended)\` options \u2014 defining UI modality & CRUD prerequisites) + ProductShape (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); \`status\` starts \`draft\`. |
20077
20114
  | \`agent-threat-modeler\` | STRIDE/OWASP threat triad (\`ACT-THREAT-*\`, \`ABUSE-*\`, \`SEC-REQ-*\`); hands off to \`agent-privacy-officer\`. |
20078
20115
  | \`agent-privacy-officer\` | LINDDUN & GDPR/LOPDGDD privacy triad (\`DATA-ASSET-*\`, \`PRIV-ROPA-*\`, \`PRIV-REQ-*\`) + Pre-Merge \`verify:privacy\`. |
20079
20116
  | \`agent-system-architect\` | Modular arc42 + NAF v4 architecture (\`CMP-*\`, \`ARCH-DEP-001\`, Mermaid diagrams, \`ADR-*\`). |
@@ -20208,7 +20245,7 @@ Non-Negotiable Quality Thresholds (\`quality-policy.yaml\`):
20208
20245
 
20209
20246
  | Role | Mission (summary) |
20210
20247
  | :--- | :--- |
20211
- | \`agent-product-analyst\` | Phase 0 NAF v4 Enterprise & Capability Vision (\`ARCH-INTRO-*\`, up to 15-question discovery interview defining UI modality, delivery model & CRUD prerequisites) + ProductShape artifacts (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); \`status\` starts \`draft\`. |
20248
+ | \`agent-product-analyst\` | Phase 0 NAF v4 Enterprise & Capability Vision (\`ARCH-INTRO-*\`, \`/grill-me\` Socratic decision-tree discovery interview of up to 15 questions \u2014 1\u20132 per turn with \`(Recommended)\` options \u2014 defining UI modality, delivery model & CRUD prerequisites) + ProductShape artifacts (\`ACT-*\`, \`UC-*\`, \`FR-*\`, \`QR-*\`, \`BR-*\`); \`status\` starts \`draft\`. |
20212
20249
  | \`agent-threat-modeler\` | STRIDE/OWASP threat triad (\`ACT-THREAT-*\`, \`ABUSE-*\`, \`SEC-REQ-*\`); hands off to \`agent-privacy-officer\`. |
20213
20250
  | \`agent-privacy-officer\` | LINDDUN & GDPR/LOPDGDD privacy triad (\`DATA-ASSET-*\`, \`PRIV-ROPA-*\`, \`PRIV-REQ-*\`) + Pre-Merge \`verify:privacy\`. |
20214
20251
  | \`agent-system-architect\` | Modular arc42 + NAF v4 architecture (\`CMP-*\`, \`ARCH-DEP-001\`, Mermaid diagrams, \`ADR-*\`). |