agent-contracts

dsl-update: description: >- DSL Update — create and update agent-contracts DSL definitions and bindings, verify quality via validate / lint / render / score. Executed by DSL Designer. entry_conditions: - A DSL definition change request exists trigger: >- Execute when new DSL definitions need to be created, existing definitions need updating, or bindings need to be added or modified. steps: - type: delegate task: update-dsl-definitions from_agent: dsl-designer description: >- DSL Designer creates or updates DSL definitions and verifies via validate / lint. - type: gate gate_kind: dsl-task-result description: >- Block if validation_result or lint_result did not pass in update-dsl-definitions. Enforces dsl-validate-before-render guardrail — prevents render from executing against invalid DSL. - type: delegate task: update-dsl-binding from_agent: dsl-designer description: >- DSL Designer adds or updates bindings and verifies via generate. - type: gate gate_kind: dsl-task-result description: >- Block if generate_result is not 'pass' in update-dsl-binding. Prevents render-dsl-outputs from executing against invalid bindings. - type: delegate task: render-dsl-outputs from_agent: dsl-designer description: >- DSL Designer renders prompts and documents and checks for drift. - type: gate gate_kind: dsl-task-result description: >- Block if render_result is not 'pass' or drift_detected is true in render-dsl-outputs result. - type: delegate task: check-dsl-score from_agent: dsl-designer description: >- DSL Designer checks completeness score and identifies improvement areas. dsl-audit: description: >- DSL Audit — audit completeness of DSL definitions against generated prompts, detect gaps, and present improvement recommendations. Executed by DSL Auditor. Supports four audit types: render (semantic diff), dsl (design coherence), prompt (prompt fidelity), and extensions (x-* consumption gap detection). entry_conditions: - DSL definition rendering is complete trigger: >- Execute when DSL completeness audit is needed. Typically run as a quality check after DSL updates. steps: - type: delegate task: audit-dsl-completeness from_agent: dsl-auditor description: >- DSL Auditor executes 19-dimension cross-check and produces audit report with improvement recommendations. - type: gate gate_kind: dsl-audit-result description: >- Block if audit-dsl-completeness detected 3 or more critical-level gaps (triggers stop_and_report escalation). - type: delegate task: audit-semantic-design from_agent: dsl-auditor depends_on: ["gate:dsl-audit-result"] description: >- DSL Auditor reviews DSL design for semantic coherence — role overlap, scope breadth, gate placement, guardrail enforcement paths. - type: gate gate_kind: dsl-audit-result depends_on: [audit-semantic-design] description: >- Semantic-design gate — block if critical design issues detected (gate placement defects, handoff schema gaps). - type: delegate task: audit-generated-prompts from_agent: dsl-auditor depends_on: ["gate:dsl-audit-result"] description: >- DSL Auditor compares generated prompts against DSL intent — detects missing requirements, hallucinated permissions, ambiguous instructions. - type: gate gate_kind: dsl-audit-result depends_on: [audit-generated-prompts] description: >- Hallucinated-permissions gate — block immediately if audit-generated-prompts detected permissions not declared in DSL (enforces dsl-no-hallucinated-permissions guardrail). - type: delegate task: audit-extension-consumption from_agent: dsl-auditor depends_on: ["gate:dsl-audit-result"] description: >- DSL Auditor checks x-* extension properties for consumption gaps — declared but unused, populated but not rendered, semantic overlap with standard DSL features. - type: gate gate_kind: dsl-audit-result depends_on: [audit-semantic-design, audit-generated-prompts, audit-extension-consumption] description: >- Terminal gate — aggregates all audit results and blocks if any critical-level findings remain unresolved.