trellis-finish-work
Pass
Audited by Gen Agent Trust Hub on Apr 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local Python scripts (
./.trellis/scripts/task.pyand./.trellis/scripts/add_session.py) to manage project tasks and journal entries. These are vendor-provided resources internal to the project structure. - [PROMPT_INJECTION]: The skill processes session metadata such as task names, titles, and summaries as command-line arguments, creating a surface for indirect prompt injection.
- Ingestion points: Arguments passed to Python scripts in
SKILL.md. - Boundary markers: Absent in the command templates.
- Capability inventory: Local command execution via shell.
- Sanitization: Absent in templates; relies on the agent's string handling and the underlying scripts' input validation.
Audit Metadata