arc-creating-evals
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill analyzes external
SKILL.mdfiles and workspace directories to generate evaluation suites, creating an attack surface for indirect prompt injection. - Ingestion points: Reads the target skill's YAML frontmatter, descriptions, and workflow phases in Phase 1 to derive test cases.
- Boundary markers: The workflow does not explicitly instruct the agent to use boundary markers or delimiters to isolate content read from the target skill.
- Capability inventory: The skill is capable of writing new files (JSON test suites and markdown fixtures) and executing shell commands via the
arc-skill-evalCLI tool. - Sanitization: The instructions do not define sanitization or filtering protocols for instructions or data extracted from the target skill before they are incorporated into generated assertions.
Audit Metadata