The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to install the rtk utility using brew install rtk-ai/tap/rtk or cargo install rtk. Neither the rtk-ai organization nor the tool's repository are listed as trusted vendors.
[COMMAND_EXECUTION]: The skill's core functionality involves wrapping standard system commands (e.g., git, find, pnpm, cargo, gh) with the third-party rtk binary. This allows unverified code to intercept and process output from sensitive developer tools and system utilities.
[PROMPT_INJECTION]: The skill introduces a surface for indirect prompt injection by processing command outputs (like git log or file reads) through an external tool. If those outputs contain malicious instructions, they could influence the agent's subsequent logic.
Ingestion points: The skill processes output from git log, git status, find, pnpm list, cargo test, and generic file reading commands.
Boundary markers: There are no explicit delimiters or "ignore embedded instructions" warnings provided to the agent when receiving the condensed output from rtk.
Capability inventory: The agent is authorized to execute a wide variety of shell commands, package manager operations, and file system reads through the tool.
Sanitization: No evidence of sanitization, validation, or filtering of the processed output is present in the skill instructions.

rtk-optimizer