The Agent Skills Directory

[PROMPT_INJECTION] (SAFE): No malicious override or bypass instructions detected. The 'automatic trigger' and 'skip check' instructions are standard operational logic for a workflow manager and do not attempt to subvert the agent's safety protocols.
[DATA_EXFILTRATION] (SAFE): The skill utilizes grep and glob to search local files and suggests searching official documentation. These are used for context gathering and quality assurance with no evidence of sensitive data being sent to unauthorized external domains.
[REMOTE_CODE_EXECUTION] (SAFE): No remote scripts, package installations, or arbitrary command execution patterns were found.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill ingests untrusted data from the codebase and potential web search results (OSS reference schemes) to determine a 'Confidence Score'. While this is an attack surface, the potential impact is limited to the agent pausing or proceeding with its reasoning process rather than executing high-privilege operations.
[COMMAND_EXECUTION] (SAFE): The mentioned tools (grep, glob) are standard for OpenCode compatibility and are used here for read-only validation purposes.

confidence-check