library-cleanup
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it utilizes the
cat_filetool to read the raw content of untrusted files to determine their purpose and generate new filenames.\n - Ingestion points: Document content read from potentially untrusted files in SKILL.md (Phase 2, Step B).\n
- Boundary markers: Absent. The instructions do not provide explicit delimiters or directives to the agent to ignore instructions or commands that might be embedded within the documents being processed.\n
- Capability inventory: The skill possesses significant capabilities including renaming files, moving files, deleting files and folders, and executing code to generate visualizations and reports.\n
- Sanitization: No sanitization or validation logic is defined for the content extracted from files before it is used to influence the agent's plan or formulated into code execution.\n- [COMMAND_EXECUTION]: The skill uses the
execute_codetool to calculate library organization metrics and generate a self-contained HTML impact dashboard. Because the data being processed—such as file names and content summaries—originates from untrusted sources, there is a risk of data-driven execution issues if the agent incorrectly interpolates these values into the scripts or the resulting HTML report.
Audit Metadata