fable-domain
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill performs 'bounded web research' in Stage 2 to identify domain-specific regulations, authorities, and failure modes. This data is then used to generate workflows, adapters, and trap fixtures. This ingestion of untrusted external content creates a surface for indirect prompt injection, where an attacker-controlled website could influence the agent's generated output.
- [COMMAND_EXECUTION]: In Stage 4, the skill executes local repository tools including a 'check script', 'fable-judge', and a 'smoke eval' harness. These tools are used to verify the integrity and behavior of the generated code bundle, which involves executing shell commands and scripts within the local environment.
- [DATA_EXPOSURE]: The skill is instructed to read all existing adapters and governing documents within the repository (e.g.,
references/domains/,SKILL.md,CHANGELOG,README) to orient itself and learn the schema. While this is necessary for its stated purpose, it demonstrates the capability to access and process the contents of the local file system.
Audit Metadata