The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (cat) and local Python scripts (python3 ./.trellis/scripts/get_context.py) to retrieve project state and documentation.
[DYNAMIC_EXECUTION]: The agent is directed to run local scripts with varying flags (--mode phase, --mode packages, etc.) to dynamically determine the workflow routing and context initialization based on the current environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from project-local files and script outputs to determine its next actions, which could be exploited if an attacker can modify the repository's configuration files.
Ingestion points: Data is read from .trellis/spec/guides/index.md, .trellis/spec/<package>/<layer>/index.md, and the console output of get_context.py.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when processing this external data.
Capability inventory: The skill has the capability to execute Python scripts, read files, and route the session to various other skills based on input content.
Sanitization: There is no evidence of sanitization or validation performed on the ingested file content or script outputs before they influence the agent's logic.

trellis-start