The Agent Skills Directory

[SAFE]: The skill does not contain any malicious patterns, obfuscation, or unauthorized access mechanisms. All identified behaviors are consistent with its stated purpose as a medical reference tool.
[PROMPT_INJECTION]: No malicious injection patterns or bypass attempts were found. The skill actively implements safety protocols, such as a 'gate check' to verify the user's medical credentials before proceeding, which helps mitigate misuse.
[DATA_EXFILTRATION]: There are no commands or scripts that access sensitive local files (e.g., .env, .ssh, or cloud credentials) or perform unauthorized network requests to external domains.
[REMOTE_CODE_EXECUTION]: The skill does not download, install, or execute external packages or scripts. It relies entirely on text-based instructions and predefined templates.
[INDIRECT_PROMPT_INJECTION]: While the skill processes user-provided data (such as drug names and patient profiles), the risk is minimal as the skill lacks capabilities to execute code, write to the filesystem, or access the network.
Ingestion points: User input provided via the /pharmacist-helper command and its arguments (e.g., drug names, patient weight).
Boundary markers: The skill uses Markdown headers and tables to structure data, though it lacks explicit instructions to ignore instructions embedded in the input strings.
Capability inventory: The skill is limited to text generation and does not utilize any system-level tools or network APIs.
Sanitization: No explicit sanitization or filtering of input data is performed beyond role-based gating.

pharmacist-helper