- •0.9.0 retains the released context repairs; independent acceptance of the historical critical findings remains pending
- •prepare_context() normalizes equivalent dotted, nested and mixed JSON; conflicts and missing required data fail closed
- •0.9.0 rechecks live authority at effects, carries pure-call restrictions, and rejects unsupported build semantics
- •Durable auditing requires an attached archive and required-audit configuration; hardened mode alone does not provision storage
- •Authority provenance and authenticated claims
- •Proof verification and signature strictness
- •Replay protection with freshness and revocation controls
- •Hash-linked audit-chain integrity
- •Capability mediation across side-effect adapters
- •Path confinement and symlink/traversal resistance
- •Trusted signed time for temporal policy checks
- •Fail-closed parsing and denial propagation
- •Governed build behavior and explicit governance-loss declarations
- •Known hardened-runtime boundaries still listed in threat model and status docs
Security · v0.9.0
Threat model and hardening surfaces
This page summarizes public threat-model documentation without publishing exploit instructions beyond already documented boundaries.
Documented Source authority: `docs/THREAT_MODEL.md` and `docs/STATUS.md` at pinned commit b09bbfc3dda4654cca2e9872423460667eed9498.
Known boundary This portal is a documentary surface; it does not execute upstream hardened runtime attack suites itself.