tfactory-fixloop
Warn
Audited by Gen Agent Trust Hub on Jun 19, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute shell commands and dynamic Python scripts. It interpolates variables such as <task_id> and <spec_dir> directly into command strings (e.g., 'cd apps/backend && python -m agents.handback <spec_dir> --send'). This pattern is vulnerable to command injection if identifiers contain shell metacharacters or malformed paths.
- [COMMAND_EXECUTION]: The skill performs dynamic code execution by generating Python scripts at runtime via shell heredocs ('python
- <<'PY''). It interpolates context variables into the script body, which creates a risk for code injection within the Python interpreter if input values are not strictly validated.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting untrusted data from external JSON reports (verdicts.json, handback_request.json) and triage documents produced by the TFactory service. Ingestion points: External findings and triage reports. Boundary markers: None identified; data is processed directly to determine control flow. Capability inventory: Includes Bash execution, file access via Read, and MCP tools for rerunning tasks and applying code corrections. Sanitization: None observed; the skill relies on direct string interpolation and JSON parsing of external data to drive its decision logic.
Audit Metadata