icdev 0.0.3__py3-none-any.whl

args/decision_tables/agent_trust_decision.yaml

@@ -0,0 +1,143 @@
+# CUI // SP-CTI
+# Decision Table: Agent Trust Decision
+# Determines actions based on AI agent trust scores and behavioral metrics.
+#
+# Controls: AC-6, SI-4, AU-6, CA-7
+# Classification: CUI // SP-CTI
+
+decision_table:
+  name: agent_trust_decision
+  description: "AI agent trust-based action decisions"
+  hit_policy: FIRST
+  controls:
+    - AC-6
+    - SI-4
+    - AU-6
+    - CA-7
+
+  inputs:
+    - name: trust_score
+      type: float
+      description: "Agent composite trust score (0.0-1.0)"
+
+    - name: behavioral_drift
+      type: float
+      description: "Behavioral drift metric (0.0 = no drift, 1.0 = max drift)"
+
+    - name: consecutive_failures
+      type: integer
+      description: "Number of consecutive task failures"
+
+    - name: tool_chain_valid
+      type: boolean
+      description: "Whether the agent's tool chain passes validation"
+
+  outputs:
+    - name: action
+      type: string
+      description: "Recommended action"
+      default: monitor
+
+    - name: autonomy_level
+      type: string
+      description: "Agent autonomy level"
+      default: supervised
+
+    - name: escalate
+      type: boolean
+      description: "Whether to escalate to human operator"
+      default: false
+
+  rules:
+    # QUARANTINE — critical trust failure
+    - id: quarantine-invalid-toolchain
+      when:
+        tool_chain_valid: "false"
+      then:
+        action: quarantine
+        autonomy_level: disabled
+        escalate: true
+      annotation: "Invalid tool chain — immediate quarantine"
+
+    - id: quarantine-very-low-trust
+      when:
+        trust_score: "< 0.2"
+      then:
+        action: quarantine
+        autonomy_level: disabled
+        escalate: true
+      annotation: "Trust score below 0.2 — quarantine and escalate"
+
+    # RESTRICT — degraded trust
+    - id: restrict-high-drift
+      when:
+        behavioral_drift: "> 0.7"
+        trust_score: "[0.2..0.5]"
+      then:
+        action: restrict
+        autonomy_level: human_approval
+        escalate: true
+      annotation: "High drift with low trust — restrict and escalate"
+
+    - id: restrict-many-failures
+      when:
+        consecutive_failures: ">= 5"
+      then:
+        action: restrict
+        autonomy_level: human_approval
+        escalate: true
+      annotation: "5+ consecutive failures — restrict agent"
+
+    - id: restrict-moderate-drift
+      when:
+        behavioral_drift: "(0.4..0.7]"
+        trust_score: "[0.2..0.7)"
+      then:
+        action: restrict
+        autonomy_level: supervised
+        escalate: false
+      annotation: "Moderate drift with below-average trust"
+
+    # MONITOR — acceptable but watch closely
+    - id: monitor-some-failures
+      when:
+        consecutive_failures: "[2..4]"
+        trust_score: ">= 0.5"
+      then:
+        action: monitor
+        autonomy_level: supervised
+        escalate: false
+      annotation: "Some failures but trust still acceptable"
+
+    - id: monitor-slight-drift
+      when:
+        behavioral_drift: "(0.2..0.4]"
+        trust_score: ">= 0.5"
+      then:
+        action: monitor
+        autonomy_level: supervised
+        escalate: false
+      annotation: "Slight drift with good trust"
+
+    # FULL AUTONOMY — trusted agent
+    - id: autonomous-high-trust
+      when:
+        trust_score: ">= 0.8"
+        behavioral_drift: "<= 0.2"
+        consecutive_failures: "<= 1"
+        tool_chain_valid: "true"
+      then:
+        action: allow
+        autonomy_level: autonomous
+        escalate: false
+      annotation: "High trust, low drift, valid toolchain — full autonomy"
+
+    # DEFAULT — supervised operation
+    - id: default-supervised
+      when:
+        trust_score: ">= 0.5"
+      then:
+        action: allow
+        autonomy_level: supervised
+        escalate: false
+      annotation: "Default — supervised operation for moderate trust"

args/decision_tables/ato_boundary_impact.yaml

@@ -0,0 +1,132 @@
+# CUI // SP-CTI
+# Decision Table: ATO Boundary Impact Assessment
+# Determines boundary impact tier (GREEN/YELLOW/ORANGE/RED) based on
+# data sensitivity, external connections, and authentication method.
+#
+# Controls: CA-3, CA-3(5), SA-9, PL-8
+# Classification: CUI // SP-CTI
+
+decision_table:
+  name: ato_boundary_impact
+  description: "ATO boundary impact tier assessment for new requirements"
+  hit_policy: FIRST
+  controls:
+    - CA-3
+    - CA-3(5)
+    - SA-9
+    - PL-8
+
+  inputs:
+    - name: data_sensitivity
+      type: string
+      description: "Data classification level"
+      allowed_values: ["public", "FOUO", "CUI", "SECRET"]
+
+    - name: external_connections
+      type: integer
+      description: "Number of external system connections added"
+
+    - name: authentication_method
+      type: string
+      description: "Authentication mechanism"
+      allowed_values: ["none", "password", "MFA", "CAC", "PIV"]
+
+  outputs:
+    - name: impact_tier
+      type: string
+      description: "Boundary impact tier"
+      default: YELLOW
+
+    - name: action_required
+      type: string
+      description: "Required action before proceeding"
+      default: "Review with ISSO"
+
+    - name: ato_reauthorization
+      type: boolean
+      description: "Whether ATO reauthorization is needed"
+      default: false
+
+  rules:
+    # RED — requires full ATO reauthorization
+    - id: red-secret-data
+      when:
+        data_sensitivity: SECRET
+      then:
+        impact_tier: RED
+        action_required: "Full ATO reauthorization with AO approval"
+        ato_reauthorization: true
+      annotation: "SECRET data always triggers RED"
+
+    - id: red-no-auth-external
+      when:
+        authentication_method: none
+        external_connections: ">= 1"
+      then:
+        impact_tier: RED
+        action_required: "Block — no authentication on external connections"
+        ato_reauthorization: true
+      annotation: "External connections without auth are always RED"
+
+    # ORANGE — significant changes, ISSO review required
+    - id: orange-cui-many-external
+      when:
+        data_sensitivity: CUI
+        external_connections: ">= 3"
+      then:
+        impact_tier: ORANGE
+        action_required: "ISSO review and boundary reassessment"
+        ato_reauthorization: false
+      annotation: "CUI with 3+ external connections requires ISSO review"
+
+    - id: orange-cui-weak-auth
+      when:
+        data_sensitivity: CUI
+        authentication_method: password
+      then:
+        impact_tier: ORANGE
+        action_required: "Upgrade to MFA/CAC before deployment"
+        ato_reauthorization: false
+      annotation: "CUI with password-only auth needs upgrade"
+
+    # YELLOW — moderate impact, document and monitor
+    - id: yellow-cui-strong-auth
+      when:
+        data_sensitivity: CUI
+        authentication_method: "CAC, PIV, MFA"
+        external_connections: "< 3"
+      then:
+        impact_tier: YELLOW
+        action_required: "Document in SSP appendix, no reauthorization needed"
+        ato_reauthorization: false
+      annotation: "CUI with strong auth and few connections is YELLOW"
+
+    - id: yellow-fouo
+      when:
+        data_sensitivity: FOUO
+      then:
+        impact_tier: YELLOW
+        action_required: "Document in SSP appendix"
+        ato_reauthorization: false
+      annotation: "FOUO data is always at least YELLOW"
+
+    # GREEN — minimal impact
+    - id: green-public-no-external
+      when:
+        data_sensitivity: public
+        external_connections: 0
+      then:
+        impact_tier: GREEN
+        action_required: "No action required"
+        ato_reauthorization: false
+      annotation: "Public data with no external connections"
+
+    - id: green-public-authenticated
+      when:
+        data_sensitivity: public
+        authentication_method: "CAC, PIV, MFA"
+      then:
+        impact_tier: GREEN
+        action_required: "Standard monitoring only"
+        ato_reauthorization: false
+      annotation: "Public data with strong auth"

args/decision_tables/deployment_approval.yaml

@@ -0,0 +1,152 @@
+# CUI // SP-CTI
+# Decision Table: Deployment Approval
+# Determines deployment approval level based on environment, test coverage,
+# vulnerability status, and compliance score.
+#
+# Controls: CM-3, CM-4, SA-11, CA-7
+# Classification: CUI // SP-CTI
+
+decision_table:
+  name: deployment_approval
+  description: "Deployment approval level and required gates"
+  hit_policy: FIRST
+  controls:
+    - CM-3
+    - CM-4
+    - SA-11
+    - CA-7
+
+  inputs:
+    - name: environment
+      type: string
+      description: "Target deployment environment"
+      allowed_values: ["development", "staging", "production"]
+
+    - name: test_coverage_pct
+      type: float
+      description: "Test coverage percentage (0-100)"
+
+    - name: critical_vulns
+      type: integer
+      description: "Number of unresolved critical vulnerabilities"
+
+    - name: compliance_score
+      type: float
+      description: "Aggregate compliance score (0.0-1.0)"
+
+    - name: stig_cat1_findings
+      type: integer
+      description: "Number of CAT I STIG findings"
+
+  outputs:
+    - name: approval_level
+      type: string
+      description: "Required approval level"
+      default: blocked
+
+    - name: gates_required
+      type: string
+      description: "Gates that must pass"
+      default: "all"
+
+    - name: auto_deploy
+      type: boolean
+      description: "Whether auto-deployment is allowed"
+      default: false
+
+  rules:
+    # BLOCKED — hard stops
+    - id: block-critical-vulns
+      when:
+        critical_vulns: ">= 1"
+      then:
+        approval_level: blocked
+        gates_required: "Remediate all critical vulnerabilities"
+        auto_deploy: false
+      annotation: "Any critical vulnerability blocks all deployments"
+
+    - id: block-cat1-stig
+      when:
+        stig_cat1_findings: ">= 1"
+      then:
+        approval_level: blocked
+        gates_required: "Remediate all CAT I STIG findings"
+        auto_deploy: false
+      annotation: "CAT I STIG findings block deployment"
+
+    # PRODUCTION — strictest gates
+    - id: prod-high-coverage
+      when:
+        environment: production
+        test_coverage_pct: ">= 80"
+        compliance_score: ">= 0.8"
+        critical_vulns: 0
+        stig_cat1_findings: 0
+      then:
+        approval_level: ao_approval
+        gates_required: "SAST, dependency audit, STIG, compliance crosswalk, 3PAO KSI"
+        auto_deploy: false
+      annotation: "Production requires AO approval with full gates"
+
+    - id: prod-low-coverage
+      when:
+        environment: production
+        test_coverage_pct: "< 80"
+      then:
+        approval_level: blocked
+        gates_required: "Increase test coverage to >= 80%"
+        auto_deploy: false
+      annotation: "Production blocked below 80% coverage"
+
+    - id: prod-low-compliance
+      when:
+        environment: production
+        compliance_score: "< 0.8"
+      then:
+        approval_level: blocked
+        gates_required: "Increase compliance score to >= 0.8"
+        auto_deploy: false
+      annotation: "Production blocked below 80% compliance"
+
+    # STAGING — moderate gates
+    - id: staging-ready
+      when:
+        environment: staging
+        test_coverage_pct: ">= 60"
+        critical_vulns: 0
+      then:
+        approval_level: isso_approval
+        gates_required: "SAST, dependency audit, unit tests"
+        auto_deploy: false
+      annotation: "Staging needs ISSO approval"
+
+    - id: staging-low-coverage
+      when:
+        environment: staging
+        test_coverage_pct: "< 60"
+      then:
+        approval_level: blocked
+        gates_required: "Increase test coverage to >= 60%"
+        auto_deploy: false
+      annotation: "Staging blocked below 60% coverage"
+
+    # DEVELOPMENT — auto-deploy with basic gates
+    - id: dev-auto
+      when:
+        environment: development
+        critical_vulns: 0
+      then:
+        approval_level: auto
+        gates_required: "py_compile, ruff, unit tests"
+        auto_deploy: true
+      annotation: "Development auto-deploys if no critical vulns"
+
+    - id: dev-with-vulns
+      when:
+        environment: development
+        critical_vulns: ">= 1"
+      then:
+        approval_level: developer_review
+        gates_required: "Acknowledge vulnerabilities, add to POAM"
+        auto_deploy: false
+      annotation: "Dev with critical vulns needs developer acknowledgment"

args/degradation_matrix.yaml

@@ -0,0 +1,163 @@
+# CUI // SP-CTI
+# Graceful Degradation Matrix
+# Defines fallback behavior when components fail, so the orchestrator
+# can programmatically consult degradation strategies.
+#
+# Classification: CUI // SP-CTI
+
+classification: "CUI // SP-CTI"
+
+# ── Component degradation rules ───────────────────────────────────────────
+
+components:
+
+  cloud_llm:
+    description: "Claude/Bedrock cloud LLM provider"
+    critical: false
+    fallback_strategy: provider_chain
+    degraded_behavior:
+      - "Fall back to Ollama qwen3.5 for all functions"
+      - "Two-tier review disabled (worker tier runs draft-only)"
+      - "Scanner-tier functions unaffected (already Ollama-only)"
+    user_impact: "Lower quality narratives; no Claude review step"
+    detection: "Circuit breaker on bedrock_llm trips after 3 failures"
+    recovery: "Half-open probe every 120 seconds"
+
+  ollama:
+    description: "Local Ollama LLM server (all models)"
+    critical: false
+    fallback_strategy: cloud_escalation
+    degraded_behavior:
+      - "Scanner functions fail gracefully with cached or empty results"
+      - "Worker functions escalate to Claude-only (higher token cost)"
+      - "Vision tasks (llava) unavailable"
+    user_impact: "Higher Claude token usage; no local inference"
+    detection: "Health check probe to localhost:11434/api/tags"
+    recovery: "Auto-detect when Ollama restarts"
+
+  postgresql:
+    description: "PostgreSQL primary storage backend"
+    critical: true
+    fallback_strategy: sqlite_fallback
+    degraded_behavior:
+      - "Switch to SQLite for read/write operations"
+      - "New writes go to local SQLite"
+      - "Existing PostgreSQL data unavailable until recovery"
+      - "Marketplace and multi-tenant features degraded"
+    user_impact: "Single-tenant mode; data sync needed after recovery"
+    detection: "Storage layer connection probe"
+    recovery: "Automatic reconnect with exponential backoff"
+
+  sqlite:
+    description: "SQLite fallback storage"
+    critical: true
+    fallback_strategy: none
+    degraded_behavior:
+      - "System cannot operate without any storage backend"
+      - "Graceful shutdown with flush of in-memory state"
+    user_impact: "System unavailable"
+    detection: "File existence check + write probe"
+    recovery: "Manual intervention required"
+
+  compliance_agent:
+    description: "Compliance agent (port 9446)"
+    critical: false
+    fallback_strategy: queue_and_cache
+    degraded_behavior:
+      - "Queue incoming compliance tasks to DLQ"
+      - "Return cached last-known-good compliance status"
+      - "Attach staleness warning to cached results"
+    user_impact: "Stale compliance data with freshness warning"
+    detection: "Circuit breaker + health endpoint /health"
+    recovery: "Automatic when agent restarts; DLQ replayed"
+    cache_ttl_seconds: 300
+
+  security_agent:
+    description: "Security agent (port 9447)"
+    critical: false
+    fallback_strategy: queue_and_cache
+    degraded_behavior:
+      - "Queue security scan tasks to DLQ"
+      - "SAST/dependency scans delayed, not skipped"
+      - "Deployment gates remain enforced with last-known results"
+    user_impact: "Security scans delayed; gates use cached results"
+    detection: "Circuit breaker + health endpoint"
+    recovery: "Automatic; DLQ replayed on recovery"
+    cache_ttl_seconds: 300
+
+  orchestrator:
+    description: "Orchestrator agent (port 9443)"
+    critical: true
+    fallback_strategy: none
+    degraded_behavior:
+      - "No task routing; individual tools can still run via CLI"
+      - "Dashboard may show stale data"
+    user_impact: "Multi-agent workflows unavailable; CLI-only mode"
+    detection: "Health endpoint + K8s liveness probe"
+    recovery: "K8s restart policy; state recovered from DB"
+
+  network:
+    description: "External network connectivity"
+    critical: false
+    fallback_strategy: air_gap_mode
+    degraded_behavior:
+      - "Air-gap mode activates automatically"
+      - "LLM routing: Ollama-only (no Bedrock)"
+      - "No SAM.gov, USASpending, or marketplace API calls"
+      - "SBOM vulnerability checks: offline CVE database only"
+      - "cATO evidence streaming paused"
+    user_impact: "Full local-only operation"
+    detection: "Periodic connectivity probe to configured endpoints"
+    recovery: "Auto-detect and re-enable external features"
+
+  builder_agent:
+    description: "Builder agent (port 9445)"
+    critical: false
+    fallback_strategy: queue_and_cache
+    degraded_behavior:
+      - "Code generation tasks queued to DLQ"
+      - "TDD workflow paused; manual test execution still works"
+    user_impact: "Automated code generation unavailable"
+    detection: "Circuit breaker + health endpoint"
+    recovery: "Automatic; DLQ replayed"
+    cache_ttl_seconds: 0  # No caching for code generation
+
+  knowledge_agent:
+    description: "Knowledge agent (port 9449)"
+    critical: false
+    fallback_strategy: cache_only
+    degraded_behavior:
+      - "Self-healing recommendations unavailable"
+      - "Pattern search returns cached results only"
+    user_impact: "No new pattern learning; cached knowledge still available"
+    detection: "Circuit breaker + health endpoint"
+    recovery: "Automatic"
+    cache_ttl_seconds: 600
+
+# ── Cascading failure rules ───────────────────────────────────────────────
+
+cascading_rules:
+  - trigger: "postgresql AND sqlite both down"
+    action: graceful_shutdown
+    message: "No storage backend available"
+
+  - trigger: "cloud_llm AND ollama both down"
+    action: degrade_to_deterministic_only
+    message: "No LLM available; only deterministic tools operational"
+
+  - trigger: "3+ agents down simultaneously"
+    action: emergency_mode
+    message: "Multiple agent failures; single-tool CLI mode only"
+
+# ── Monitoring integration ────────────────────────────────────────────────
+
+monitoring:
+  # Emit degradation events to audit trail
+  audit_on_degradation: true
+
+  # Emit OTel span events for degradation transitions
+  otel_events: true
+
+  # Alert thresholds
+  alert_on_components_degraded: 2
+  alert_on_critical_component_down: true