valent-pipeline 0.2.19 → 0.2.21

package/pipeline/steps/reqs/library.md

@@ -0,0 +1,56 @@
+# REQS Step: Library Requirements
+
+## Step 4b: Library-Specific Requirement Extraction
+
+For stories with `library` in `testing_profiles`, extract additional requirements:
+
+### Public API Surface
+- Exported symbols (functions, classes, constants, types)
+- Function signatures with parameter types and return types
+- Overload variants (if applicable)
+- Re-export paths (subpath exports via package.json `exports` map)
+
+### Semver Policy
+- What constitutes a breaking change (removed export, changed signature, changed behavior)
+- What constitutes a minor change (new export, new optional parameter)
+- What constitutes a patch change (bug fix, internal refactor, docs)
+- Current version and expected version after this story ships
+
+### Backwards Compatibility Constraints
+- Previously published API surface that must remain stable
+- Deprecated symbols and their migration path
+- Minimum supported runtime versions (Node.js, browser targets)
+- Compatibility with existing consumer code patterns
+
+### Module System Requirements
+- Package format: CJS, ESM, or dual
+- Entry points: `main`, `module`, `exports` map configuration
+- Conditional exports (node, import, require, default)
+- Build output structure and file extensions (.js, .mjs, .cjs)
+
+### Type Declaration Requirements
+- Type declaration strategy: `.d.ts` files, inline TypeScript, JSDoc
+- Generic type parameters and constraints
+- Strict mode compatibility (noImplicitAny, strictNullChecks)
+- Type export completeness (every public symbol has a corresponding type)
+
+### Peer Dependency Contract
+- Required peer dependencies with version ranges
+- Optional peer dependencies and behavior when absent
+- Conflict resolution for overlapping dependency ranges
+- Runtime detection of missing peers (clear error messages)
+
+### Tree-Shaking Requirements
+- `sideEffects` field value (false, or list of side-effectful files)
+- Barrel re-export avoidance (no index files that re-export everything)
+- Per-export isolation (importing one symbol must not pull in unrelated code)
+- Bundle size expectations for selective imports
+
+## Quality Gate Additions
+
+- [ ] Every exported symbol has defined behavior and type
+- [ ] Semver impact classified (breaking, minor, or patch)
+- [ ] Module system entry points specified (CJS, ESM, or dual)
+- [ ] Type declarations defined for every public export
+- [ ] Peer dependency contract documented with version ranges
+- [ ] Tree-shaking requirements specified if sideEffects: false is declared

package/pipeline/steps/reqs/mcp-server.md

@@ -0,0 +1,48 @@
+# REQS Step: MCP Server Requirements
+
+## Step 4b: MCP-Server-Specific Requirement Extraction
+
+For stories with `mcp-server` in `testing_profiles`, extract additional requirements:
+
+### Tool Definitions
+- Tool name, human-readable description, and purpose
+- Which tools the server exposes at `tools/list`
+- Tool grouping or categorization (if server exposes many tools)
+
+### Input Schemas per Tool
+- Required and optional parameters with types
+- Parameter constraints (min/max, regex, enum values, string length)
+- Nested object schemas and array item types
+- Default values for optional parameters
+
+### Output Content Types per Tool
+- Content type per tool response (text, image, audio, resource)
+- Response shape: key fields, MIME types, encoding
+- Multi-content responses (tools returning mixed content types)
+- Resource URIs and their resolution behavior
+
+### Capability Declarations
+- What the server advertises in `initialize` response (tools, resources, prompts)
+- Server info (name, version)
+- Protocol version compatibility
+- Resource subscription support (if applicable)
+
+### Transport Requirements
+- Supported transport(s): stdio, SSE, streamable HTTP
+- Connection lifecycle (startup, shutdown, reconnect behavior)
+- Concurrency model (serial tool calls vs parallel)
+- Timeout behavior per tool (long-running operations, cancellation)
+
+### Error Model per Tool
+- JSON-RPC error codes for protocol-level failures (-32700, -32600, -32601, -32602, -32603)
+- `isError: true` content responses for tool-level failures (business logic errors)
+- Specific error scenarios per tool (invalid input, resource not found, permission denied, rate limited)
+- Error message format and detail level
+
+## Quality Gate Additions
+
+- [ ] Every tool has explicit error behavior defined (both JSON-RPC codes and isError:true scenarios)
+- [ ] Every tool has an input schema with required/optional params and types
+- [ ] Every tool has output content type(s) specified
+- [ ] Capability declarations documented for initialize handshake
+- [ ] Transport requirements specified

package/pipeline/steps/reqs/mobile-app.md

@@ -0,0 +1,54 @@
+# REQS Step: Mobile App Requirements
+
+## Step 4b: Mobile-App-Specific Requirement Extraction
+
+For stories with `mobile-app` in `testing_profiles`, extract additional requirements:
+
+### Screen Inventory
+- Screen names and navigation hierarchy (stack, tab, drawer, modal)
+- Deep link URI patterns per screen (e.g., `myapp://settings/profile`)
+- Back button / gesture behavior per screen (Android hardware back, iOS swipe-to-go-back)
+- Screen transition animations (if specified in UX)
+
+### Platform-Specific Behavior
+- Android-only vs iOS-only vs shared behavior per AC
+- Permission requirements per platform (camera, location, notifications, storage, contacts, biometrics)
+- Minimum OS version targets (Android API level, iOS version)
+- Platform-specific UI patterns (Material Design vs Human Interface Guidelines)
+
+### Offline and Connectivity Requirements
+- Offline-capable features and sync behavior (queue-and-retry, optimistic UI, conflict resolution)
+- Network error handling per screen (retry, fallback, offline banner)
+- Local storage / caching strategy (AsyncStorage, MMKV, SQLite, Hive)
+- Background sync requirements
+
+### Device Form Factor
+- Phone vs tablet layout requirements
+- Orientation support (portrait, landscape, both) per screen
+- Safe area / notch / dynamic island handling
+- Keyboard avoidance behavior for input screens
+
+### Push Notification Requirements
+- Notification types and payload structure
+- Foreground vs background handling behavior
+- Deep link from notification tap (target screen, parameters)
+- Silent / data-only notification handling
+
+### App Lifecycle Requirements
+- Background → foreground behavior (refresh data, re-authenticate)
+- App termination recovery (restore navigation state, pending form data)
+- Multi-window / split-screen behavior (if applicable)
+
+### Native Module Dependencies
+- Camera, biometrics, file system, Bluetooth, NFC, etc.
+- Third-party SDK requirements (maps, analytics, crash reporting)
+- Platform-specific native module availability
+
+## Quality Gate Additions
+
+- [ ] Every screen has navigation path and deep link documented
+- [ ] Platform-specific behavior explicit for each AC (or explicitly marked "identical on both platforms")
+- [ ] Permission requirements listed per platform
+- [ ] Offline behavior specified (or explicitly marked as online-only)
+- [ ] Push notification handling defined (or explicitly marked as no-push)
+- [ ] App lifecycle edge cases addressed (background/foreground, termination recovery)

package/pipeline/steps/reqs/self-review.md

@@ -10,13 +10,15 @@ Verify completeness:
 - [ ] Every AC has a priority classification (P0-P3)
 - [ ] All ambiguities are documented with resolution and rationale
 - [ ] Business rules cover all ACs with edge case notes
-- [ ] Database changes are specified (or explicitly marked "none" with rationale)
-- [ ] API endpoints are specified (or explicitly marked "none" with rationale)
-- [ ] Auth requirements are defined
+- [ ] If api or data-pipeline in testing_profiles: database changes are specified (or explicitly marked "none" with rationale)
+- [ ] If api in testing_profiles: API endpoints are specified (or explicitly marked "none" with rationale)
+- [ ] If api or ui in testing_profiles: auth requirements are defined
 - [ ] Cross-cutting concerns are addressed
 - [ ] NFR targets are set
 - [ ] Pre-mortem has at least 3 findings
 - [ ] All pre-mortem `folded-in` findings are reflected in the brief
 - [ ] Cross-references section points to all consumed inputs
+- [ ] All quality gate additions from loaded domain step files (Step 4b) are satisfied
+- [ ] Domain-specific conditional sections from template are populated for each active profile
 
 If any check fails, go back and fix it before proceeding.

package/pipeline/task-graphs/backend-api.yaml

@@ -37,12 +37,20 @@ tasks:
     activeForm: "BEND implementing backend"
     blockedBy: [readiness]
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [bend]
+    blockedBy: [bend, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -102,12 +110,21 @@ agents:
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
     cd_targets: [bend]
 
+  - name: IAC
+    ref: iac
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/iac.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [iac]
+
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from BEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]
 

package/pipeline/task-graphs/data-pipeline.yaml

@@ -30,19 +30,27 @@ tasks:
     activeForm: "READINESS reviewing pre-dev artifacts"
     blockedBy: [qa_a]
 
-  - ref: bend
-    agent: BEND
-    subject: "BEND: Implement data pipeline production code and tests"
-    description: "Read reqs-brief.md and qa-test-spec.md, implement data pipeline, produce bend-handoff.md."
-    activeForm: "BEND implementing data pipeline"
+  - ref: data
+    agent: DATA
+    subject: "DATA: Implement data pipeline production code and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement data pipeline, produce data-handoff.md."
+    activeForm: "DATA implementing data pipeline"
     blockedBy: [readiness]
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [bend]
+    blockedBy: [data, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -89,25 +97,34 @@ agents:
     wave: 1
     prompt: .valent-pipeline/prompts/readiness.md
     trigger: "Wait for [HANDOFF] from QA-A. Begin when task is unblocked."
-    completion: "Send [READINESS-APPROVAL] to BEND, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
+    completion: "Send [READINESS-APPROVAL] to DATA, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
     cd_targets: [readiness]
 
   # Wave 2: Spawn when QA-A completes (spin up while READINESS reviews)
-  - name: BEND
-    ref: bend
+  - name: DATA
+    ref: data
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/data.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [data]
+
+  - name: IAC
+    ref: iac
     wave: 2
     spawn_trigger: "qa_a completed"
-    prompt: .valent-pipeline/prompts/bend.md
+    prompt: .valent-pipeline/prompts/iac.md
     trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
-    cd_targets: [bend]
+    cd_targets: [iac]
 
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from BEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]
 

package/pipeline/task-graphs/document-generation.yaml

@@ -30,19 +30,27 @@ tasks:
     activeForm: "READINESS reviewing pre-dev artifacts"
     blockedBy: [qa_a]
 
-  - ref: bend
-    agent: BEND
-    subject: "BEND: Implement template and content pipeline production code and tests"
-    description: "Read reqs-brief.md and qa-test-spec.md, implement document generation templates and content pipeline, produce bend-handoff.md."
-    activeForm: "BEND implementing document generation"
+  - ref: docgen
+    agent: DOCGEN
+    subject: "DOCGEN: Implement template and content pipeline production code and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement document generation templates and content pipeline, produce docgen-handoff.md."
+    activeForm: "DOCGEN implementing document generation"
     blockedBy: [readiness]
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [bend]
+    blockedBy: [docgen, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -89,25 +97,34 @@ agents:
     wave: 1
     prompt: .valent-pipeline/prompts/readiness.md
     trigger: "Wait for [HANDOFF] from QA-A. Begin when task is unblocked."
-    completion: "Send [READINESS-APPROVAL] to BEND, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
+    completion: "Send [READINESS-APPROVAL] to DOCGEN, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
     cd_targets: [readiness]
 
   # Wave 2: Spawn when QA-A completes (spin up while READINESS reviews)
-  - name: BEND
-    ref: bend
+  - name: DOCGEN
+    ref: docgen
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/docgen.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [docgen]
+
+  - name: IAC
+    ref: iac
     wave: 2
     spawn_trigger: "qa_a completed"
-    prompt: .valent-pipeline/prompts/bend.md
+    prompt: .valent-pipeline/prompts/iac.md
     trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
-    cd_targets: [bend]
+    cd_targets: [iac]
 
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from BEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]
 

package/pipeline/task-graphs/frontend-only.yaml

@@ -51,12 +51,20 @@ tasks:
     activeForm: "FEND implementing frontend"
     blockedBy: [readiness]
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [fend]
+    blockedBy: [fend, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -132,12 +140,21 @@ agents:
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
     cd_targets: [fend]
 
+  - name: IAC
+    ref: iac
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/iac.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [iac]
+
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from FEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]
 

package/pipeline/task-graphs/fullstack-web.yaml

@@ -59,12 +59,20 @@ tasks:
     blockedBy: [readiness]
     skip_when: "testing_profiles excludes ui"
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [bend, fend]
+    blockedBy: [bend, fend, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -149,12 +157,21 @@ agents:
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
     cd_targets: [fend]
 
+  - name: IAC
+    ref: iac
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/iac.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [iac]
+
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from both BEND and FEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]
 

package/pipeline/task-graphs/library.yaml

@@ -30,19 +30,27 @@ tasks:
     activeForm: "READINESS reviewing pre-dev artifacts"
     blockedBy: [qa_a]
 
-  - ref: bend
-    agent: BEND
-    subject: "BEND: Implement public API and module production code and tests"
-    description: "Read reqs-brief.md and qa-test-spec.md, implement library public API and module, produce bend-handoff.md."
-    activeForm: "BEND implementing library"
+  - ref: libdev
+    agent: LIBDEV
+    subject: "LIBDEV: Implement public API and module production code and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement library public API and module, produce libdev-handoff.md."
+    activeForm: "LIBDEV implementing library"
     blockedBy: [readiness]
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [bend]
+    blockedBy: [libdev, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -89,25 +97,34 @@ agents:
     wave: 1
     prompt: .valent-pipeline/prompts/readiness.md
     trigger: "Wait for [HANDOFF] from QA-A. Begin when task is unblocked."
-    completion: "Send [READINESS-APPROVAL] to BEND, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
+    completion: "Send [READINESS-APPROVAL] to LIBDEV, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
     cd_targets: [readiness]
 
   # Wave 2: Spawn when QA-A completes (spin up while READINESS reviews)
-  - name: BEND
-    ref: bend
+  - name: LIBDEV
+    ref: libdev
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/libdev.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [libdev]
+
+  - name: IAC
+    ref: iac
     wave: 2
     spawn_trigger: "qa_a completed"
-    prompt: .valent-pipeline/prompts/bend.md
+    prompt: .valent-pipeline/prompts/iac.md
     trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
-    cd_targets: [bend]
+    cd_targets: [iac]
 
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from BEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]
 

package/pipeline/task-graphs/mcp-server.yaml

@@ -30,19 +30,27 @@ tasks:
     activeForm: "READINESS reviewing pre-dev artifacts"
     blockedBy: [qa_a]
 
-  - ref: bend
-    agent: BEND
-    subject: "BEND: Implement MCP tool production code and tests"
-    description: "Read reqs-brief.md and qa-test-spec.md, implement MCP server tools and handlers, produce bend-handoff.md."
-    activeForm: "BEND implementing MCP server"
+  - ref: mcp_dev
+    agent: MCP-DEV
+    subject: "MCP-DEV: Implement MCP tool production code and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement MCP server tools and handlers, produce mcp-dev-handoff.md."
+    activeForm: "MCP-DEV implementing MCP server"
     blockedBy: [readiness]
 
+  - ref: iac
+    agent: IAC
+    subject: "IAC: Implement infrastructure definitions and tests"
+    description: "Read reqs-brief.md and qa-test-spec.md, implement infrastructure code, produce iac-handoff.md."
+    activeForm: "IAC implementing infrastructure"
+    blockedBy: [readiness]
+    conditional: "testing_profiles includes iac"
+
   - ref: critic
     agent: CRITIC
     subject: "CRITIC: Adversarial code review"
     description: "Read git-diff, reqs-brief.md, qa-test-spec.md. Run blind-hunt, edge-case-hunt, acceptance-audit, triage passes. Produce critic-review.md."
     activeForm: "CRITIC reviewing code"
-    blockedBy: [bend]
+    blockedBy: [mcp_dev, iac]
 
   - ref: qa_b
     agent: QA-B
@@ -89,25 +97,34 @@ agents:
     wave: 1
     prompt: .valent-pipeline/prompts/readiness.md
     trigger: "Wait for [HANDOFF] from QA-A. Begin when task is unblocked."
-    completion: "Send [READINESS-APPROVAL] to BEND, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
+    completion: "Send [READINESS-APPROVAL] to MCP-DEV, or [READINESS-REJECTION] to responsible agent. Mark task completed on approval."
     cd_targets: [readiness]
 
   # Wave 2: Spawn when QA-A completes (spin up while READINESS reviews)
-  - name: BEND
-    ref: bend
+  - name: MCP-DEV
+    ref: mcp_dev
+    wave: 2
+    spawn_trigger: "qa_a completed"
+    prompt: .valent-pipeline/prompts/mcp-dev.md
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
+    completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
+    cd_targets: [mcp-dev]
+
+  - name: IAC
+    ref: iac
     wave: 2
     spawn_trigger: "qa_a completed"
-    prompt: .valent-pipeline/prompts/bend.md
+    prompt: .valent-pipeline/prompts/iac.md
     trigger: "Begin immediately — you were spawned because QA-A completed. Wait for READINESS approval before starting implementation."
     completion: "Send [HANDOFF] to Lead and CRITIC. Mark task completed."
-    cd_targets: [bend]
+    cd_targets: [iac]
 
   - name: CRITIC
     ref: critic
     wave: 2
     spawn_trigger: "qa_a completed"
     prompt: .valent-pipeline/prompts/critic.md
-    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from BEND before starting review."
+    trigger: "Begin immediately — you were spawned because QA-A completed. Wait for [HANDOFF] from all active dev agents before starting review."
     completion: "Send [HANDOFF] to Lead and QA-B. Mark task completed."
     cd_targets: [critic]