The Agent Skills Directory

[SAFE]: The skill serves as a documentation resource providing orchestration patterns and dispatch guidance for AI subagents. It defines a structured approach for decomposing tasks and managing context isolation without introducing malicious instructions.- [INDIRECT_PROMPT_INJECTION]: The skill explicitly addresses the risk of indirect prompt injection through URL-embedded content in tool outputs. It provides mitigation strategies, such as stripping URL fragments/anchors and using schema gates for untrusted inputs to prevent the agent from following malicious instructions hidden in external data.- [REMOTE_CODE_EXECUTION]: The documentation mentions the creation of 'init.sh' files and the execution of build/test scripts. These are described as standard components of a coordinated developer workflow for managing state across sessions or verifying code artifacts, and do not involve fetching or executing untrusted code from remote sources.- [COMMAND_EXECUTION]: The skill references the use of tools like 'Bash' for running tests, lints, and builds in a verifier role. This is a standard and intended use case for development-oriented agents, and the skill emphasizes using these tools for deterministic checks before applying LLM judgment.

agent