Core model
Deterministic authorization before execution
Every automated action passes policy evaluation before irreversible work. No receipt, no execution. Fail-closed by default, if authorization cannot be established, the action is blocked.
Block unauthorized actions at the execution boundary, before payments, deployments, or data mutations reach production.
- Pre-execution PERMIT / DENY / SILENCE / ESCALATE
- Fail-closed when gate unreachable
- Evaluation separated from execution
Every decision produces a cryptographically signed receipt, verifiable offline against published keys, independent of TrigGuard uptime.
- Ed25519-signed receipts
- Key discovery via well-known URI
- Replayable policy evaluation
Evidence survives audit, tamper-evident records that do not require TrigGuard to vouch after the fact.
- Immutable decision records
- Minimal disclosure, metadata only
- No hidden policy layers
Category thesis
Why authorization is a security control
Traditional systems monitor execution. TrigGuard governs execution. The difference is infrastructure, not tooling.
- Detect
- Investigate
- Respond
- Recover
- Evaluate
- Decide
- Block
- Prove
Architecture
High-level components
Evaluation separated from execution, the TG-01 authorization boundary.
Policy surfaces
Configuration and policy bundles you integrate with.
Execution gate
Deterministic PERMIT / DENY / SILENCE / ESCALATE before commit.
Policy evaluation
Rules evaluated against declared action metadata.
Receipt infrastructure
Ed25519-signed decision records over canonical payload.
Verification authority
Offline checks via /.well-known/trigguard-keys.json.
Site practices
HTTPS, HSTS, CSP on hosted pages. Metadata-only evaluation by design.
Evidence
Proof, not promises
What the system commits to under declared policy, verifiable, not interpretive.
Fail-closed
No authority established, execution blocked.
Cryptographic receipts
Ed25519-signed, tamper-evident decision records.
Offline verification
Validate receipts without TrigGuard uptime.
Deterministic decisions
Same input, same policy, same outcome every time.
Policy enforcement
Declared rules bound to actions, not narratives.
Evidence chain
Audit-ready proof from intent to outcome. Verify a receipt →
Responsible disclosure
Security disclosure program
We welcome good-faith reports affecting the website, protocol interfaces, verification flows, or hosted infrastructure.
48-hour acknowledgement
Initial response target for reports to security@trigguardai.com.
Coordinated disclosure
Good-faith researchers who avoid privacy violations and service disruption.
Security contact
Abuse contact
security.txt
RFC 9116 well-known location. View file →
- POST /execute
- /.well-known/trigguard-keys.json
- Receipt verification
- Site vulnerabilities
- Hosted authorization runtime
No public bug bounty program at this time. Include reproduction steps, affected endpoints, and impact assessment when reporting.
Cryptography
Protocol cryptography
Ed25519 over canonicalized JSON, immutable once signed, verifiable offline.
Ed25519 - live conformance key
/.well-known/trigguard-keys.json (JWK Set, Ed25519 / EdDSA)
Lexicographic key sort, recursive, undefined-omitted, no whitespace
Optional payload_hash = hex SHA-256 of canonical signed payload
Additive-only; new algorithms may be added without removing Ed25519
Compliance
Standards compliance
security.txt (RFC 9116)
Security contact information published at the standard well-known location:
Contact: mailto:security@trigguardai.com Contact: mailto:abuse@trigguardai.com Expires: 2027-01-01T00:00:00.000Z Preferred-Languages: en Canonical: https://www.trigguardai.com/.well-known/security.txt Policy: https://www.trigguardai.com/security/.well-known/security.txt