ljg-roundtable
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill explicitly directs the agent to execute the shell command
date +%Y%m%dT%H%M%Sin the '存档' (Archive) section of SKILL.md to generate a timestamp for file naming. - [INDIRECT_PROMPT_INJECTION]: The archiving logic incorporates user-supplied debate topics into a local file path:
~/Context/{timestamp}--圆桌-{议题关键词}__roundtable.org. The instructions lack requirements for the agent to sanitize the{议题关键词}variable, which could allow a malicious user to provide path traversal sequences (e.g.,../../) to influence the write location of the log file. - Ingestion points: User-provided debate topic used to derive keywords in SKILL.md.
- Boundary markers: None present; the instructions direct the agent to use the user topic directly.
- Capability inventory: The skill requires file system write capabilities and shell command execution (
date) to complete its workflow. - Sanitization: No sanitization or validation steps are defined for user-provided strings used in terminal commands or file path construction.
Audit Metadata