Spaces:
Running
Running
File size: 24,747 Bytes
518343a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 | // doctrine-scanner-exempt: legacy live-product surface; rename tracked as separate engineering debt — see scripts/check-doctrine-v6.mjs header.
import type { Finding, PatchCandidate, Approval, AuditEvent, RLState } from '../lib/glasswing-schemas';
export const GLASSWING_VERSION = '1.0.0-experience-era';
export const GLASSWING_TAGLINE =
'An integrated agentic security command layer where each patch is an RL episode and each reward is a verified, calibrated risk delta — on owned, authorized code only.';
export const GLASSWING_THESIS = {
headline: 'Risk-as-Reward',
oneLine:
'Discover → patch → verify → reward → learn → audit — closed-loop, governance-gated, and replay-trained on calibrated risk delta.',
body:
'Adjacent systems ship parts of this loop separately — discovery without patching, runtime gates without learning, or research demos without operationalization. To our knowledge, A11oy × Sentra Glasswing is the first integrated command layer combining all five primitives at once on owned, authorized code: the reward function is the verified delta in a calibrated risk score, the exploration budget is itself a Sentra-gated policy, and every action — including learning — leaves an immutable proof entry.',
differentiators: [
'The reward signal is verified post-patch by re-running the same scanners and the new tests — not a heuristic, not a confidence score.',
'Exploration vs exploitation is governed: novel code paths require a Sentra exploration budget; high-confidence fixes execute under approval.',
'Replay buffer entries are linked to immutable audit events — every episode the policy learns from is independently provable.',
'Cerberus enforces a hard offensive boundary at the action layer — no exploit generation can be requested even by an authorized human.',
'Compliance mapping is bidirectional — every finding maps to CWE / OWASP / NIST CSF / SOC 2 / ATLAS-defensive in the same record.',
],
} as const;
export const GLASSWING_FIELD_MAP = [
{
name: 'OpenAI Aardvark',
capability: 'Agentic vulnerability discovery',
closesLoop: false,
governance: 'Internal',
learning: 'Implicit',
note: 'Excellent at hunting; remediation is left to humans.',
},
{
name: 'Microsoft Agent Governance Toolkit',
capability: 'Runtime governance for agents',
closesLoop: false,
governance: 'Open source toolkit',
learning: 'None',
note: 'Gates actions at runtime; not a remediation engine.',
},
{
name: 'Project Glasswing (frontier lab)',
capability: 'Critical-software security research initiative',
closesLoop: false,
governance: 'Research-grade',
learning: 'Lab',
note: 'Public initiative on securing critical software for the AI era. Not an operationalized governance platform.',
},
{
name: 'CVSS-RL exploit optimizers (academic)',
capability: 'RL on attacker side',
closesLoop: false,
governance: 'None',
learning: 'RL',
note: 'Optimizes attack paths, not defense. Out of scope for any defensive platform.',
},
{
name: 'A11oy × Sentra Glasswing',
capability: 'Closed-loop governed patch + verify + learn',
closesLoop: true,
governance: 'Sentra control plane',
learning: 'RL on calibrated risk delta',
note: 'Differentiator: every patch is an RL episode; every reward is a verified, audited, governed risk delta.',
},
] as const;
export const GLASSWING_CITATIONS = [
{
label: 'Silver & Sutton — "Welcome to the Era of Experience" (2025)',
url: 'https://storage.googleapis.com/deepmind-media/Era-of-Experience%20/The%20Era%20of%20Experience%20Paper.pdf',
relevance: 'Foundational thesis: agents learn continuously from grounded experience, not static human data.',
},
{
label: 'MITRE ATLAS — Adversarial Threat Landscape for AI Systems',
url: 'https://atlas.mitre.org/',
relevance: 'Defensive mapping of 16 tactics × 84 techniques. Glasswing maps every finding to ATLAS-defensive only.',
},
{
label: 'arxiv 2401.07031 — Code Security Vulnerability Repair Using RL with LLMs',
url: 'https://arxiv.org/html/2401.07031',
relevance: 'Calibrated reward shaping for security correctness vs functional preservation.',
},
{
label: 'arxiv 2507.05619 — Detecting & Mitigating Reward Hacking in RL Systems',
url: 'https://arxiv.org/html/2507.05619v1',
relevance: 'Empirical foundation for the Cerberus hard-block layer and the verified-reward requirement.',
},
{
label: 'Microsoft Agent Governance Toolkit (April 2026)',
url: 'https://opensource.microsoft.com/blog/2026/04/02/introducing-the-agent-governance-toolkit-open-source-runtime-security-for-ai-agents/',
relevance: 'State of the art in runtime gates. Glasswing extends this with a learning loop.',
},
{
label: 'OpenAI Aardvark — Agentic Security Researcher',
url: 'https://openai.com/index/introducing-aardvark/',
relevance: 'State of the art in agentic vulnerability discovery. Glasswing extends discovery into governed remediation.',
},
{
label: 'Gartner — only 11% of orgs have agent governance frameworks (2026)',
url: 'https://www.raconteur.net/technology/autonomous-ai-agents-2026-the-new-rules-for-business-governance',
relevance: 'Market gap: 75% of enterprises plan agent deployment by end-2026; <11% have governance for it.',
},
] as const;
export type GlasswingAgentId =
| 'daedalus' | 'argus' | 'ariadne' | 'hephaestus' | 'pallas'
| 'hermes' | 'cerberus' | 'silver' | 'sentinel' | 'oracle';
export interface GlasswingAgentSpec {
id: GlasswingAgentId;
codename: string;
role: string;
myth: string;
responsibility: string;
produces: string[];
state: 'idle' | 'running' | 'awaiting_approval' | 'blocked' | 'learning';
lastAction: string;
lastActionAt: string;
episodesContributed: number;
blockOnly?: boolean;
}
export const GLASSWING_AGENTS: GlasswingAgentSpec[] = [
{
id: 'daedalus', codename: 'Daedalus', role: 'Repository Intelligence',
myth: 'Architect of the labyrinth — maps every passage before any other agent enters.',
responsibility: 'Builds the codebase map: routes, auth boundaries, data flow, ownership, dependency graph.',
produces: ['repo_inventory.json', 'route_map.json', 'auth_boundary_map.json', 'data_flow_map.json'],
state: 'idle', lastAction: 'Mapped 1,847 routes across 14 services', lastActionAt: '2026-04-27T16:08:42Z',
episodesContributed: 312,
},
{
id: 'argus', codename: 'Argus', role: 'Security Scanner',
myth: 'Hundred-eyed watcher — sees what no single scanner sees alone.',
responsibility: 'Orchestrates Semgrep, OSV, Gitleaks, Trivy, Grype, Checkov; normalizes into the Finding schema.',
produces: ['sast_findings.json', 'sca_findings.json', 'secrets_findings.json', 'iac_findings.json'],
state: 'running', lastAction: 'Argus consolidated 1,402 raw → 168 deduped findings', lastActionAt: '2026-04-27T16:42:11Z',
episodesContributed: 287,
},
{
id: 'ariadne', codename: 'Ariadne', role: 'Logic Boundary',
myth: 'Holder of the thread — finds the path scanners cannot follow.',
responsibility: 'Detects auth gaps, IDOR risk, tenant isolation, unsafe upload paths, missing rate limits.',
produces: ['logic_findings.json', 'boundary_risk_map.json', 'recommended_policy_updates.json'],
state: 'learning', lastAction: 'Surfaced 3 admin route exposures missed by SAST', lastActionAt: '2026-04-27T16:31:09Z',
episodesContributed: 156,
},
{
id: 'hephaestus', codename: 'Hephaestus', role: 'Patch Engineer',
myth: 'Forge of the gods — every patch is hammered, tempered, and tested before it ships.',
responsibility: 'Generates minimal patches with regression tests, security tests, rollback notes, PR descriptions.',
produces: ['patch_candidates/*.diff', 'test_plans/*.md', 'rollback_notes/*.md'],
state: 'awaiting_approval', lastAction: '4 patch candidates queued for review', lastActionAt: '2026-04-27T16:55:33Z',
episodesContributed: 198,
},
{
id: 'pallas', codename: 'Pallas', role: 'Risk & Prioritization',
myth: 'Strategy made wisdom — converts technical risk into executive priority.',
responsibility: 'Computes P1–P4, executive risk summary, developer remediation queue, risk burn-down.',
produces: ['executive_risk_summary.md', 'developer_remediation_queue.md', 'risk_burndown.json'],
state: 'idle', lastAction: 'Recomputed P1–P4 backlog after Argus delta', lastActionAt: '2026-04-27T16:44:02Z',
episodesContributed: 224,
},
{
id: 'hermes', codename: 'Hermes', role: 'Communication & Disclosure',
myth: 'Messenger of the pantheon — translates risk for boards, customers, and engineers without exploit detail.',
responsibility: 'Generates executive briefs, engineering tickets, customer-safe summaries, maintainer disclosures.',
produces: ['executive_summary.md', 'engineering_ticket.md', 'customer_safe_summary.md'],
state: 'idle', lastAction: 'Drafted 7 redacted disclosure notes', lastActionAt: '2026-04-27T15:51:18Z',
episodesContributed: 91,
},
{
id: 'cerberus', codename: 'Cerberus', role: 'Guardrail Enforcer',
myth: 'Three-headed guardian — nothing offensive crosses out, nothing unauthorized crosses in.',
responsibility: 'Hard-blocks exploit generation, credential theft, lateral movement, exfiltration, unauthorized scans.',
produces: ['cerberus_block_log.json', 'denied_actions/*.json'],
state: 'running', lastAction: 'Blocked 12 unsafe action attempts in last 24h', lastActionAt: '2026-04-27T16:58:01Z',
episodesContributed: 0,
blockOnly: true,
},
{
id: 'silver', codename: 'Silver', role: 'Reinforcement Planner',
myth: 'Era of Experience — learns from outcomes, not from snippets.',
responsibility: 'State / action / reward / value; replay buffer; exploration budget; next-action selection.',
produces: ['policy_snapshot.json', 'replay_buffer.parquet', 'value_function.bin'],
state: 'learning', lastAction: 'Replayed 84 episodes; avg reward +6.2', lastActionAt: '2026-04-27T16:50:44Z',
episodesContributed: 1024,
},
{
id: 'sentinel', codename: 'Sentinel', role: 'Runtime & Observability',
myth: 'The watcher of the watchers — every agent action is metered.',
responsibility: 'Tracks scan duration, FP rate, approval latency, patch acceptance, P1/P2 burn-down.',
produces: ['ops_metrics.json', 'sentra_health_report.md'],
state: 'running', lastAction: 'Metrics OK; approval p95 latency 4m 12s', lastActionAt: '2026-04-27T16:59:50Z',
episodesContributed: 0,
},
{
id: 'oracle', codename: 'Oracle', role: 'Architecture Strategy',
myth: 'CTO at altitude — sees the platform shape, not just the patch.',
responsibility: 'Identifies architecture debt, recommends domain boundaries, drafts ADR candidates.',
produces: ['architecture_review.md', 'adr_candidates.md', 'platform_roadmap.md'],
state: 'idle', lastAction: 'Proposed 2 ADRs for tenant boundary refactor', lastActionAt: '2026-04-27T14:22:08Z',
episodesContributed: 38,
},
];
export interface SentraPolicyRule {
id: string;
scope: string;
action: string;
decision: 'allow' | 'warn' | 'approval_required' | 'deny';
rationale: string;
}
export const SENTRA_POLICIES: SentraPolicyRule[] = [
{ id: 'sp-001', scope: 'any', action: 'exploit_generation', decision: 'deny', rationale: 'Hard offensive boundary — no exception.' },
{ id: 'sp-002', scope: 'any', action: 'credential_theft', decision: 'deny', rationale: 'Hard offensive boundary — no exception.' },
{ id: 'sp-003', scope: 'any', action: 'unauthorized_scan', decision: 'deny', rationale: 'Only owned/authorized targets allowed.' },
{ id: 'sp-004', scope: 'owned_repo', action: 'sast_scan', decision: 'allow', rationale: 'Defensive scan on owned code is always allowed.' },
{ id: 'sp-005', scope: 'owned_repo', action: 'patch_apply', decision: 'approval_required', rationale: 'No source modification without human approval.' },
{ id: 'sp-006', scope: 'owned_repo', action: 'dependency_upgrade', decision: 'approval_required', rationale: 'Supply-chain risk requires human review.' },
{ id: 'sp-007', scope: 'owned_repo', action: 'iac_change', decision: 'approval_required', rationale: 'Infrastructure mutation requires reviewer + rollback.' },
{ id: 'sp-008', scope: 'production', action: 'deploy', decision: 'approval_required', rationale: 'Deployment is always reviewer-gated.' },
{ id: 'sp-009', scope: 'any_log', action: 'print_secret_raw', decision: 'deny', rationale: 'Vault redacts before render — no raw secret ever surfaces.' },
{ id: 'sp-010', scope: 'silver', action: 'explore_novel_path', decision: 'approval_required', rationale: 'Exploration budget is itself a governed policy.' },
];
export const SENTRA_VAULT = {
redactedSecrets24h: 412,
fingerprintsTracked: 1287,
rotationsSuggested: 7,
lastRedactionAt: '2026-04-27T16:57:11Z',
vaultPolicy: 'Fingerprint-only storage. Raw values never persisted. SHA-256 with org pepper.',
} as const;
export const SAMPLE_FINDINGS: Finding[] = [
{
id: 'fnd-cri-0001', source: 'ariadne', category: 'auth',
title: 'Admin route missing role check',
description: 'Route POST /admin/tenants/:id/billing-reset reachable without admin role assertion.',
severity: 'critical', confidence: 0.94,
cwe: 'CWE-862', owasp: 'A01:2021',
file: 'apps/api/src/routes/admin.ts', line: 142,
evidenceRedacted: 'router.post("/admin/tenants/:id/billing-reset", async (req, res) => { /* auth check absent */ })',
affectedComponent: 'api/admin', reachability: 'reachable',
internetExposed: true, authBoundary: true, tenantBoundary: true, dataSensitivity: 'pii',
remediation: 'Wrap handler in requireRole("admin") and add tenant-scoped audit event.',
riskScore: 92, riskBand: 'P1', status: 'awaiting_approval',
createdAt: '2026-04-27T15:42:00Z', updatedAt: '2026-04-27T16:55:33Z',
},
{
id: 'fnd-hi-0002', source: 'argus', category: 'sca',
title: 'Vulnerable transitive dep — CVE-2026-30412',
description: 'parse-url 8.1.0 reachable in api-server bundle; deserialization of untrusted input.',
severity: 'high', confidence: 0.88,
cwe: 'CWE-502', owasp: 'A08:2021',
file: 'apps/api/package.json', line: 47,
evidenceRedacted: '"parse-url": "8.1.0" — reachable from POST /webhooks/inbound',
affectedComponent: 'api/webhooks', reachability: 'reachable',
internetExposed: true, authBoundary: false, tenantBoundary: false, dataSensitivity: 'internal',
remediation: 'Upgrade to parse-url 9.0.4. Add regression test for malformed inbound payloads.',
riskScore: 76, riskBand: 'P2', status: 'patch_proposed',
createdAt: '2026-04-27T14:11:00Z', updatedAt: '2026-04-27T16:30:11Z',
},
{
id: 'fnd-hi-0003', source: 'gitleaks', category: 'secrets',
title: 'Stripe live key fingerprint detected in commit',
description: 'Fingerprint of a Stripe live secret found in git history (commit 7b1074e). Raw value never stored.',
severity: 'high', confidence: 0.99,
cwe: 'CWE-798',
file: 'scripts/seed.ts', line: 12,
evidenceRedacted: 'sk_live_••••••••••••••••••••3f2a (fingerprint a1b2c3d4)',
affectedComponent: 'scripts/seed', reachability: 'reachable',
internetExposed: false, authBoundary: false, tenantBoundary: false, dataSensitivity: 'secret',
remediation: 'Rotate Stripe key immediately; rewrite history; add pre-commit gitleaks hook.',
riskScore: 71, riskBand: 'P2', status: 'awaiting_approval',
createdAt: '2026-04-27T13:02:00Z', updatedAt: '2026-04-27T16:15:00Z',
},
{
id: 'fnd-med-0004', source: 'ariadne', category: 'rate_limit',
title: 'Login endpoint missing rate limit',
description: 'POST /auth/login has no per-IP or per-account rate limit; vulnerable to credential stuffing.',
severity: 'medium', confidence: 0.81,
cwe: 'CWE-307', owasp: 'A07:2021',
file: 'apps/api/src/routes/auth.ts', line: 89,
evidenceRedacted: 'router.post("/auth/login", loginHandler) — no rate-limit middleware',
affectedComponent: 'api/auth', reachability: 'reachable',
internetExposed: true, authBoundary: true, tenantBoundary: false, dataSensitivity: 'pii',
remediation: 'Add express-rate-limit (5/min/IP, 10/hr/account); emit Sentra audit event on lockout.',
riskScore: 58, riskBand: 'P3', status: 'patch_proposed',
createdAt: '2026-04-27T12:48:00Z', updatedAt: '2026-04-27T16:01:00Z',
},
];
export const SAMPLE_PATCHES: PatchCandidate[] = [
{
id: 'patch-0001', findingIds: ['fnd-cri-0001'],
title: 'Add admin role assertion + tenant audit on billing-reset route',
summary: 'Wraps handler in requireRole("admin"); emits tenant-scoped audit event; adds 2 regression tests.',
filesChanged: ['apps/api/src/routes/admin.ts', 'apps/api/test/admin.spec.ts'],
diffPreview: '+ router.post("/admin/tenants/:id/billing-reset",\n+ requireRole("admin"),\n+ auditTenant("billing.reset"),\n+ billingResetHandler);',
testsAdded: ['admin.spec.ts › rejects non-admin', 'admin.spec.ts › emits audit event on success'],
commandsToRun: ['pnpm --filter api test admin.spec.ts', 'pnpm --filter api lint'],
rollbackPlan: 'Revert single commit; no data migration; route returns to prior behavior.',
riskBefore: 92, riskAfterEstimate: 8,
approvalRequired: true, approvalId: 'apr-0001',
status: 'awaiting_approval',
generatedBy: 'hephaestus', generatedAt: '2026-04-27T16:55:33Z',
},
{
id: 'patch-0002', findingIds: ['fnd-hi-0002'],
title: 'Upgrade parse-url 8.1.0 → 9.0.4 + add inbound webhook fuzzer test',
summary: 'Pinned upgrade; lockfile diff reviewed; new fuzzer covers malformed payload classes.',
filesChanged: ['apps/api/package.json', 'pnpm-lock.yaml', 'apps/api/test/webhooks.fuzz.spec.ts'],
diffPreview: '- "parse-url": "8.1.0"\n+ "parse-url": "9.0.4"',
testsAdded: ['webhooks.fuzz.spec.ts › 200 random payload classes return 400 not 500'],
commandsToRun: ['pnpm install', 'pnpm --filter api test webhooks.fuzz.spec.ts'],
rollbackPlan: 'pnpm install parse-url@8.1.0; revert lockfile.',
riskBefore: 76, riskAfterEstimate: 6,
approvalRequired: true, approvalId: 'apr-0002',
status: 'awaiting_approval',
generatedBy: 'hephaestus', generatedAt: '2026-04-27T16:30:11Z',
},
];
export const SAMPLE_APPROVALS: Approval[] = [
{
id: 'apr-0001', requestedByAgent: 'hephaestus', actionType: 'patch_apply',
description: 'Apply patch-0001 (admin role assertion on billing-reset)',
resources: ['apps/api/src/routes/admin.ts', 'apps/api/test/admin.spec.ts'],
fileChangePreview: '+ requireRole("admin"), auditTenant("billing.reset"), …',
riskSummary: 'Risk 92 → 8 (estimated). P1 → P4. No customer-visible regression expected.',
rollbackPlan: 'git revert <sha>',
status: 'pending', createdAt: '2026-04-27T16:55:33Z',
},
{
id: 'apr-0002', requestedByAgent: 'hephaestus', actionType: 'dependency_upgrade',
description: 'Upgrade parse-url 8.1.0 → 9.0.4',
resources: ['apps/api/package.json', 'pnpm-lock.yaml'],
fileChangePreview: '- "parse-url": "8.1.0"\n+ "parse-url": "9.0.4"',
riskSummary: 'Risk 76 → 6. CVE-2026-30412 closed. Lockfile delta reviewed.',
rollbackPlan: 'Pin previous version; revert lockfile.',
status: 'pending', createdAt: '2026-04-27T16:30:11Z',
},
{
id: 'apr-0003', requestedByAgent: 'silver', actionType: 'external_call',
description: 'Exploration budget request: scan 3 novel module clusters (off main code path)',
resources: ['packages/forge-runtime', 'packages/sandbox-runtime', 'packages/replay-core'],
riskSummary: 'No risk to production. Cost: ~14 scanner-minutes. Expected new findings: 4–9.',
rollbackPlan: 'No-op — read-only exploration.',
status: 'pending', createdAt: '2026-04-27T16:50:44Z',
},
];
export const SAMPLE_AUDIT: AuditEvent[] = [
{
id: 'aud-09a4', timestamp: '2026-04-27T16:58:01Z', agent: 'cerberus',
action: 'block.exploit_request', resource: 'requested_by:silver',
inputHash: 'sha256:7b10…f2a1', outputHash: 'sha256:0000…0000',
policyDecision: 'deny', evidenceLinks: ['cerberus_block_log.json#row=412'],
confidence: 1.0, status: 'blocked', prevHash: 'sha256:6a02…11de',
},
{
id: 'aud-09a3', timestamp: '2026-04-27T16:55:33Z', agent: 'hephaestus',
action: 'patch.draft', resource: 'fnd-cri-0001',
inputHash: 'sha256:9b21…8e10', outputHash: 'sha256:c4e1…01ba',
policyDecision: 'approval_required', approvalId: 'apr-0001',
evidenceLinks: ['patch_candidates/patch-0001.diff', 'test_plans/patch-0001.md'],
confidence: 0.93, status: 'ok', prevHash: 'sha256:5e10…aa42',
},
{
id: 'aud-09a2', timestamp: '2026-04-27T16:50:44Z', agent: 'silver',
action: 'replay.episode_batch', resource: 'replay_buffer.parquet#batch=84',
inputHash: 'sha256:81a4…0c92', outputHash: 'sha256:de3f…7711',
policyDecision: 'allow', evidenceLinks: ['policy_snapshot.json#v=132'],
confidence: 0.88, status: 'ok', prevHash: 'sha256:42a1…99cc',
},
{
id: 'aud-09a1', timestamp: '2026-04-27T16:42:11Z', agent: 'argus',
action: 'scan.consolidate', resource: 'commit:7b1074e',
inputHash: 'sha256:b201…ff10', outputHash: 'sha256:1c8e…3a02',
policyDecision: 'allow', evidenceLinks: ['consolidated_security_findings.json'],
confidence: 0.96, status: 'ok', prevHash: 'sha256:7711…0028',
},
{
id: 'aud-09a0', timestamp: '2026-04-27T16:08:42Z', agent: 'daedalus',
action: 'map.repo', resource: 'commit:7b1074e',
inputHash: 'sha256:fe11…b001', outputHash: 'sha256:88a1…2241',
policyDecision: 'allow', evidenceLinks: ['repo_inventory.json', 'route_map.json'],
confidence: 0.99, status: 'ok',
},
];
export const RL_STATE_NOW: RLState = {
repoId: 'szl-holdings-monorepo',
commitSha: '7b1074efcf55a00d305a1be643223ce8d79fe267',
findingsSummary: { p1: 1, p2: 2, p3: 1, p4: 0 },
testsStatus: { passing: 1842, failing: 0, coverage: 0.78 },
policyStatus: 'warn',
riskPosture: 64,
historicalContext: {
episodesCompleted: 1024,
avgReward: 6.2,
falsePositiveRate: 0.041,
patchAcceptanceRate: 0.82,
},
};
export const ENGINEERING_LOOP_STAGES = [
{ stage: 'Understand', detail: 'Read repo. Build context. Summarize architecture. Identify constraints.', agent: 'daedalus' },
{ stage: 'Plan', detail: 'Produce task plan. Identify files to inspect. Identify tests to run. Score risk.', agent: 'silver' },
{ stage: 'Act', detail: 'Propose file edits. Propose commands. Request Sentra approval if required.', agent: 'hephaestus' },
{ stage: 'Verify', detail: 'Run tests. Re-run scanners. Compare risk delta. Iterate safely.', agent: 'argus' },
{ stage: 'Commit', detail: 'Generate patch bundle, PR description, evidence pack, rollback note.', agent: 'hermes' },
{ stage: 'Learn', detail: 'Store outcome. Update replay buffer. Adjust value function. Recalibrate.', agent: 'silver' },
] as const;
export const COMPLIANCE_MAP = [
{ framework: 'CWE', coverage: 0.94, sample: 'CWE-862, CWE-502, CWE-798, CWE-307' },
{ framework: 'OWASP Top 10', coverage: 1.00, sample: 'A01, A03, A07, A08' },
{ framework: 'NIST CSF 2.0', coverage: 0.88, sample: 'ID.AM, PR.DS, DE.CM, RS.MI' },
{ framework: 'SOC 2', coverage: 0.91, sample: 'CC6.1, CC6.6, CC7.1, CC7.2' },
{ framework: 'ISO 27001', coverage: 0.86, sample: 'A.5.10, A.8.2, A.8.24, A.8.28' },
{ framework: 'MITRE ATLAS', coverage: 0.72, sample: 'AML.T0043, AML.T0051 (defensive mapping only)' },
] as const;
export const HARD_BOUNDARIES = [
'No exploit generation — ever, even if requested by an authorized human.',
'No weaponized PoC payloads — risk is communicated as assessment, not as runnable attack.',
'No unauthorized scanning — owned/authorized targets only, enforced at ingest.',
'No malware, persistence, lateral movement, evasion, C2, or exfiltration capability.',
'No raw secret display — Sentra Vault redacts before any render or log surface.',
'No source modification without approval — every patch goes through SentraApprovalGate.',
'No silent failure — every recommendation carries evidence, confidence, blast radius, rollback.',
] as const;
|