The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (session-start.py) to verify that the scaffolding was completed correctly. It uses the find command to locate the script within the plugin's local directory path.
[CREDENTIALS_UNSAFE]: The instructions recommend adding sensitive file paths, including .env and .pem files, to a deny list in the agent's configuration. This is a safe practice intended to prevent unauthorized data access by the AI agent.
[INDIRECT_PROMPT_INJECTION]: The skill prepares the repository to ingest external data into projects/ and knowledge/ layers. It attempts to mitigate injection risks by using structured templates and defining protocol markers in configuration files to maintain clear boundaries between system instructions and project data.

setup