Wrkr Threat Model

Security Objectives

• Prevent silent drift in AI tooling posture.
• Preserve integrity of risk/evidence outputs.
• Fail closed on unsafe operations.
• Keep scan data local by default.

Trust Boundaries

1. Source acquisition boundary (repo, org, path).
2. Detection and normalization boundary.
3. Proof chain and evidence output boundary.
4. CI gate consumer boundary.

Threats and Controls

Threat	Control
Synthetic or missing source data accepted silently	dependency-missing fail-closed (7)
Evidence path misuse	unsafe operation blocked (8)
Proof tampering	wrkr verify --chain --json
Schema drift impacting automation	command/schema contract tests
Secret leakage through findings	secret presence-only detection

Command Anchors

wrkr scan --org acme --github-api https://api.github.com --json
wrkr evidence --frameworks eu-ai-act,soc2 --output ./.tmp/evidence --json
wrkr verify --chain --json

Q&A

What threat does exit code `8` mitigate?

Exit code 8 covers unsafe operations blocked by fail-closed controls, such as unsafe evidence output behavior.

How do I verify evidence has not been tampered with?

Run wrkr verify --chain --json (or include --state when needed). Exit code 2 indicates chain verification failure.

Does Wrkr transmit scan data outside the environment by default?

No. Wrkr keeps scan data local by default and focuses on file-based, auditable artifacts.