The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents how to authorize and execute shell commands using the claude CLI's tool-calling capabilities, specifically facilitating Bash commands for git and npm tasks.
[PROMPT_INJECTION]: The skill describes a method for piping external data, such as git diff output, into the agent context via stdin. This establishes an ingestion point for untrusted data. Boundary markers are deferred to an 'active workspace authority contract' that governs external actions. The capability inventory includes Read, Grep, Glob, and Bash shell access. No specific sanitization or filtering of the interpolated data is described within the skill instructions.

claude-cli-runtime