The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides instructions for the agent to use the local cp command to initialize project-specific configuration and deferral files from internal templates. These operations are limited to local filesystem management.
[INDIRECT_PROMPT_INJECTION]: The skill is configured by reading external project files.
Ingestion points: The agent reads .architecture/config.yml to determine its assessment logic and intensity levels.
Boundary markers: None are defined to differentiate configuration data from potential instructions.
Capability inventory: Tool access is restricted to Read and Edit operations as specified in the frontmatter.
Sanitization: The skill does not perform validation or sanitization on the configuration file content.

pragmatic-guard