The Agent Skills Directory

[SAFE]: The skill instructions and documentation focus entirely on the legitimate use of the kotlinx-benchmark toolkit. No malicious patterns or security bypass attempts were detected.
[COMMAND_EXECUTION]: The script scripts/inspect-project.py performs static analysis of local project files to detect configuration settings. It uses standard library functions and does not engage in network activity, unauthorized file access, or remote code execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes project configuration data. Ingestion points: Build and source files in the project directory are read by the inspection script. Boundary markers: The script returns structured summaries to the agent. Capability inventory: The script is limited to reading project metadata and reporting status. Sanitization: Extraction is performed using specific regular expressions for benchmark-related identifiers, which limits the influence of arbitrary file content on the agent context.

kotlinx-benchmark