fable-judge
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data such as code files, diffs, and external reports in
SKILL.md, which presents an attack surface for indirect instructions.\n - Ingestion points: Reads code files, agent reports, and conversation history (
SKILL.md).\n - Boundary markers: No explicit delimiters or instructions to ignore embedded content are specified.\n
- Capability inventory: Executes shell commands and runs arbitrary scripts/tests (
SKILL.md).\n - Sanitization: No sanitization or escaping mechanisms are described for analyzed content.\n- [DYNAMIC_EXECUTION]: The skill is instructed to execute tests, builds, and scripts associated with the work being judged, and it runs logic from
eval/workflow.jsas part of its verification suite.\n- [COMMAND_EXECUTION]: Uses system commands likegit diffandgit statusto establish ground truth for file changes in the environment.\n- [EXTERNAL_DOWNLOADS]: The skill mentions cloning a repository (https://github.com/Sahir619/fable-method) to retrieve verification scenarios and workflows when not installed as a plugin.
Audit Metadata