The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or data exfiltration attempts were identified in the skill. The instructions and commands are consistent with the stated purpose of performing local model evaluations.
[COMMAND_EXECUTION]: The skill executes local shell scripts (scripts/run_local_ollama_eval.sh and scripts/run_stage2_harness.sh) and uses standard testing tools (pytest). These are legitimate actions within a local development environment.
[COMMAND_EXECUTION]: Uses chmod +x to grant execution permissions to local scripts, which is a standard step for preparing scripts for execution in a Linux/macOS environment.
[COMMAND_EXECUTION]: Performs network requests to http://localhost:11434, which is the default local endpoint for the Ollama service. This is a safe operation as it targets a local service on the user's machine.

local-multimodal-eval