The Agent Skills Directory

[SAFE]: The skill is a legitimate tool for toxicological assessment in drug discovery. It uses predefined SMARTS patterns to identify hazardous structural motifs in chemical data.
[EXTERNAL_DOWNLOADS]: The skill requires the RDKit library (rdkit). RDKit is a widely used, open-source cheminformatics toolkit maintained by a reputable community and is considered a standard dependency in the scientific domain.
[COMMAND_EXECUTION]: The instructions include standard Python CLI commands for execution (python scripts/main.py) and syntax validation (python -m py_compile). No dangerous shell features, privilege escalation, or persistence mechanisms were detected.
[DATA_EXFILTRATION]: The skill operates entirely on provided SMILES strings and outputs results to the console or workspace. No network calls (curl, wget, socket) or exfiltration patterns are present in the code or instructions.
[INDIRECT_PROMPT_INJECTION]: The skill ingests SMILES strings, which are highly structured chemical identifiers. These are parsed by the RDKit library as molecular graphs, posing negligible risk of containing adversarial natural language instructions that could subvert the agent's behavior.

toxicity-structure-alert