The Agent Skills Directory

[SAFE]: The skill instructions and configuration files contain no evidence of malicious patterns, obfuscation, or unauthorized data access.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided arguments and repository content to brainstorm options. While it lacks explicit output delimiters for these external inputs, this is a standard design for a brainstorming tool and does not grant the agent any dangerous capabilities by itself. Evidence: Ingestion occurs via $ARGUMENTS in agents/openai.yaml and repository inspection in SKILL.md; boundary markers and explicit sanitization are absent; no dangerous capabilities (like shell execution or network exfiltration) were found in the provided files.

ork-brainstorm