The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the bash tool to interact with a local script at ./.trellis/scripts/task.py. This tool is used to automate task creation and subtask management within the user's repository structure.
[DYNAMIC_EXECUTION]: The skill utilizes dynamic shell command construction where arguments such as the task title () and slug () are derived directly from user input. This pattern creates a potential command injection surface if the agent does not properly sanitize or escape these strings before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user messages and existing repository files (code, documents, and configurations) to generate PRDs and technical research.
Ingestion points: Information is gathered from user prompts and repository files via the trellis-brainstorm agent and its trellis-research sub-agent.
Boundary markers: The instructions lack explicit requirements for using delimiters or "ignore instructions" warnings when processing potentially adversarial content from the repository.
Capability inventory: The agent has capabilities for shell command execution (bash), file writing (PRD seeding), and network operations (via WebFetch/WebSearch tools in research mode).
Sanitization: There are no instructions provided to sanitize, filter, or validate the external content before it is interpolated into the PRD or shell command arguments.

trellis-brainstorm