eval-suite
Warn
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill constructs shell commands using string interpolation of user-provided content and labels (e.g.,
python scripts/eval-runner.py --text "{content_or_path}"). This pattern is vulnerable to command injection if the user input contains shell metacharacters like semicolons or backticks, allowing arbitrary code execution. - [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection by ingesting and processing untrusted content from external files and directories.
- Ingestion points: Content sources provided by the user in Step 2 and Step 3 of SKILL.md.
- Boundary markers: Absent; there are no delimiters or instructions to ignore embedded prompts in the ingested text.
- Capability inventory: Subprocess calls to
scripts/eval-runner.pyandscripts/quality-tracker.py(referenced in SKILL.md) which perform file reads and log results. - Sanitization: Absent; the skill does not specify any validation or sanitization of the processed text.
Audit Metadata