The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the Bash tool to execute several bundled Python scripts (chemistry_facts.py, crystal_validator.py, stereochem_tracker.py, smiles_verifier.py, degrees_of_unsaturation.py, molecular_formula.py). These scripts perform localized chemical calculations such as stoichiometry, density verification, and stereochemistry tracking, which is a safe and effective way to ensure numerical accuracy.
[INDIRECT_PROMPT_INJECTION]: The skill has a potential attack surface as it processes external, user-provided chemical data (molecular formulas, SMILES strings, spectroscopy results) which are then used as parameters for local script execution and external database queries.
Ingestion points: Data enters the agent's context through user chemistry queries described in SKILL.md.
Boundary markers: The instructions do not define specific delimiters for untrusted chemical data.
Capability inventory: The skill utilizes the Bash tool to run local Python scripts and several MCP tools for querying external scientific databases including PubChem, ChEMBL, HMDB, and OPSIN.
Sanitization: The bundled scripts perform basic input validation (e.g., regex parsing for formulas) but do not implement comprehensive sanitization against adversarial instruction embedding.
[SAFE]: The skill references well-known scientific services and APIs (PubChem, ChEMBL, HMDB, OPSIN) for retrieving chemical properties and metadata. These are established sources in the scientific community and their use for data retrieval is standard and safe practice.

tooluniverse-organic-chemistry