The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to analyze external data sources such as profile samples, execution traces, query plans, and runtime metrics. This creates a surface for indirect prompt injection where malicious instructions could be embedded in the performance data. 1. Ingestion points: Data entering the agent context includes profile samples, traces, allocation/lock data, and query plans as described in SKILL.md. 2. Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when the agent reads external diagnostic outputs. 3. Capability inventory: The agent has the capability to modify repository code and execute build/test commands as part of the optimization process. 4. Sanitization: There are no specific instructions for sanitizing or validating diagnostic files.
[SAFE]: The skill focuses on documentation and methodology for performance tuning. It does not contain any hardcoded credentials, remote code execution patterns, or attempts to exfiltrate data. The instructions emphasize reproducibility and systematic change management, which serves as a procedural safeguard against unintended behavior.

performance-optimization