The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the execution of bundled Python scripts (install_agent_workspace_rules.py and validate_agent_workspace.py) to set up and verify the project workspace. These scripts perform file system operations to install markdown templates and check link integrity.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest 'Source Material' from untrusted external sources, such as URLs, attachments, and pasted text, to generate or update Agent Documents. This creates a surface for malicious instructions to be introduced into the agent's context.
Ingestion points: Untrusted 'Source Material' as defined in TERMS.md and processed via workflows in agent-file-ingestion.md and context-initialization.md.
Boundary markers: Instructions suggest separating imported claims from agent-added clarifications, but strict technical delimiters are not enforced.
Capability inventory: The agent uses local scripts to write files and perform validation within the .agents/ directory.
Sanitization: Relies on manual agent review and rewording instructions rather than automated filtering.
[DYNAMIC_EXECUTION]: The installation script modifies file permissions at runtime to make the validation script executable.
Evidence: scripts/install_agent_workspace_rules.py uses chmod(0o755) on the target validation script after copying it to the workspace.

chrisai-knowledge