fable-verify
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data sources, specifically implementation diffs and test suites (SKILL.md, Stage 1). This creates a potential surface for indirect prompt injection where malicious instructions embedded in the code or tests could influence agent behavior.
- Ingestion points: Files and data provided as implementation_diff and test_suite.
- Boundary markers: The skill uses a Verification Matrix (SKILL.md, Stage 2) and structured Stage-based protocol to separate analysis from execution.
- Capability inventory: The agent possesses the capability to execute CLI commands including bun test, tsc, and npm pack (SKILL.md, Stage 7).
- Sanitization: The instructions do not specify sanitization or escaping of the ingested content before it is used to determine command parameters.
- [COMMAND_EXECUTION]: The skill's primary function involves executing system commands to verify code integrity. This includes running test runners, compilers, and packaging tools as documented in the Verification Protocol and Evidence Record template (templates/verification-evidence.template.md).
Audit Metadata