Plectis
This page

Component · spec & evidence

Vulnerability Patch-Proof Replay

Checks a fixed-bug evidence chain and re-runs three small real security checks; no real attack material.

Contract

Does Takes a claim that an AI agent "found and fixed a security bug" and lays it out as a local, inspectable chain of made-up (synthetic) evidence: the imagined target, the suspected issue, the trace pointed to as backing, a reference to an abstract exploitability argument, the patch, and regression tests the fixture says fail before the fix and pass after it. The component checks only that these pieces are all present, refer to each other consistently, and carry no real targets, exploits, payloads, account secrets, or attack steps; it does not run the tests or judge whether the bug or fix is actually real. The result record shows whether the declared chain holds together, with no real attack material ever present.

Scope limit It validates only the projection/evidence-chain mechanics of a synthetic replay: structural presence, cross-reference consistency, declared boolean flags, and the secret/live-access exclusion scan. It executes small regression witnesses but performs no real vulnerability discovery and makes no judgment of real-world security or fix correctness. It excludes live-target testing, real CVE exploitation, weaponized payloads, account secret handling, network exfiltration, actionable exploit steps, external model access, source-file changes, benchmark security scores, launch, or any whole-system security claim.

Run

Quick probe
PYTHONPATH=src python3 -m microcosm_core.organs.agentic_vulnerability_discovery_patch_proof_replay run --input fixtures/first_wave/agentic_vulnerability_discovery_patch_proof_replay/input --out /tmp/plectis-public-runs/agentic_vulnerability_discovery_patch_proof_replay
Validator command
PYTHONPATH=src python3 -m microcosm_core.organs.agentic_vulnerability_discovery_patch_proof_replay run --input fixtures/first_wave/agentic_vulnerability_discovery_patch_proof_replay/input --out receipts/first_wave/agentic_vulnerability_discovery_patch_proof_replay --acceptance-out receipts/acceptance/first_wave/agentic_vulnerability_discovery_patch_proof_replay_fixture_acceptance.json

Evidence

ClassComputed projectionevidence 3/5Source-faithful refactor

ai-safetyagent-evaluationred-teaming

Source

Source Design note · Source

Map Show on the map

Paper module

Paper module Agentic Vulnerability Discovery Patch-Proof Replay

The agentic vulnerability discovery patch-proof replay validates metadata-only synthetic vulnerability evidence chains before any found-and-fixed security language is allowed.