plotloom-asset-selection
Pass
Audited by Gen Agent Trust Hub on Oct 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to evaluate untrusted media assets (images and videos) and associated prompt intents, which serves as a potential vector for indirect prompt injection attacks where instructions hidden in the media could influence the agent's decision-making.\n
- Ingestion points: Video and image files stored within the
candidates/directory, along with user-provided prompt intent strings.\n - Boundary markers: The skill lacks explicit markers or delimiters to isolate data from instruction context.\n
- Capability inventory: The agent is authorized to execute the
plotloomCLI tool to copy and backup files.\n - Sanitization: No content sanitization or path validation steps are defined for the input assets.\n- [COMMAND_EXECUTION]: The skill instructions include the execution of the
plotloomcommand-line utility for asset selection and repository management.\n - Evidence: The
SKILL.mdfile explicitly calls for the use ofplotloom --repo <repo> select <candidate>following an 'accept' decision.
Audit Metadata