The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (UI strings, documentation, and prose) provided by users or read from repositories. This presents a potential surface for indirect prompt injection if the source text contains instructions intended to manipulate the agent's behavior. The skill mitigates this by enforcing strict preservation of placeholders, code identifiers, and structural markers, and by employing a linguistic rubric focused on fidelity to the source meaning.
[SAFE]: Analysis of the skill instructions and resource files found no evidence of network exfiltration, hardcoded credentials, or suspicious file system access. The skill operates within the local filesystem and codebase scope.
[SAFE]: No obfuscation, dynamic code execution, or privilege escalation patterns were detected. The skill relies on static language profiles and native agent search/read capabilities.

oma-translation