The Agent Skills Directory

[SAFE]: All files were analyzed for potential security threats. The content consists of professional-grade system prompts and documentation for agent orchestration. No evidence of prompt injection, obfuscation, or malicious intent was found.
[COMMAND_EXECUTION]: The 'improve-agent.md' and 'multi-agent-optimize.md' files contain documentation referencing conceptual or platform-specific tools like 'context-manager' and 'parallel-test-runner'. The only shell script provided, 'test.sh', performs a simple file-existence check and contains no dangerous commands.
[INDIRECT_PROMPT_INJECTION]: The skill defines workflows for analyzing agent logs and user feedback, which represents a data ingestion surface. However, the skill provides only instructional templates and lacks the autonomous capabilities (such as automated script execution or unvalidated file writes) necessary to exploit this surface.
[DYNAMIC_EXECUTION]: Illustrative Python code blocks are present in the 'multi-agent-optimize.md' file for demonstration purposes. These blocks are contained within markdown and are not part of an automated execution pipeline.

agent-orchestration