The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of instruction override or safety bypass attempts. The skill explicitly instructs the agent to maintain ethical boundaries and refuse deceptive requests, such as impersonation or defamation.
[DATA_EXFILTRATION]: The skill does not perform network operations or access sensitive filesystem locations (e.g., SSH keys, AWS credentials). It only interacts with its own internal roster storage and standard project logging files.
[REMOTE_CODE_EXECUTION]: There are no patterns involving external script downloads (curl | bash), package manager usage, or the execution of untrusted code. The logic is entirely contained within the provided markdown and reference files.
[DYNAMIC_EXECUTION]: No dynamic code generation or unsafe deserialization (e.g., eval, pickle) was detected. The skill's functionality is limited to natural language processing and text generation.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or private secrets were found within the skill files or metadata.
[OBFUSCATION]: The content is clear and readable. No techniques such as Base64 encoding of commands, zero-width characters, or homoglyph attacks were used to hide malicious intent.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-supplied figures and problems, it implements robust boundary markers and an 'Ethics Gate'. Given its lack of dangerous system capabilities (e.g., file writing or network access), the risk of exploiting the agent via external data is negligible.

summon