The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of the copilot CLI tool with broad administrative and filesystem permissions. Specifically, it encourages the use of flags such as --allow-all-tools, --allow-all-paths, and --yolo (as seen in SKILL.md and references/cli-command-reference.md), which grant the underlying model the ability to perform unconstrained actions on the host system.
[PROMPT_INJECTION]: The skill establishes a significant surface for Indirect Prompt Injection by forwarding untrusted data to a secondary AI model with high execution privileges.
Ingestion points: Untrusted data enters the context via user-provided task descriptions and project file contents as specified in the prompt template in SKILL.md.
Boundary markers: The skill lacks effective boundary markers or safety instructions to prevent the delegated model from interpreting data within the 'Task' or 'Context' fields as direct commands.
Capability inventory: The delegated copilot CLI possesses extensive capabilities including shell command execution and filesystem access, particularly when combined with the suggested permission flags.
Sanitization: No mechanisms are described for sanitizing or escaping external content before it is interpolated into the CLI command prompt.

copilot-cli