The Agent Skills Directory

[SAFE]: Comprehensive review of the skill's instructions and reference files revealed no security vulnerabilities or malicious intent.
[PROMPT_INJECTION]: The skill uses clear, logical workflows to guide the agent's behavior. There are no attempts to bypass safety filters, jailbreak the model, or extract system instructions.
[DATA_EXFILTRATION]: No network tools or sensitive file path access were detected. The skill mandates confidentiality and the use of the least sensitive information in its evidence-and-safety.md file.
[COMMAND_EXECUTION]: The skill does not include any shell commands, subprocess calls, or dynamic code execution patterns.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute external scripts or packages.
[CREDENTIALS_UNSAFE]: No hardcoded secrets, API keys, or private tokens were found in the codebase or metadata.

startup-operator