trellis-before-dev
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and adopt instructions from external Markdown files located in the
./.trellis/spec/directory. If these files are maliciously crafted, they could influence the agent's coding behavior or cause it to bypass security constraints during the development task. - Ingestion points:
SKILL.mdincludes instructions for reading specification and guideline files usingcatcommands. - Boundary markers: Absent. The agent is encouraged to internalize the content as mandatory development standards before proceeding to write code.
- Capability inventory: The agent uses this information to guide its primary code generation and implementation tasks.
- Sanitization: None. The content of the files is treated as authoritative project context.
- [COMMAND_EXECUTION]: The skill runs a local script (
python3 ./.trellis/scripts/get_context.py) to discover project packages. While typical for project-specific tooling, this involves executing code that resides in the local filesystem and is not directly visible in the skill definition.
Audit Metadata