The Agent Skills Directory

[SAFE]: The skill provides legitimate functionality for managing scientific configuration files in the Proteina-Complexa project. All instructions and tools described are consistent with the stated purpose of defining protein design targets.
[COMMAND_EXECUTION]: The skill utilizes the project's internal complexa CLI and standard shell utilities (grep, awk, sort) for routine tasks like listing target entries, validating configurations, and inspecting structure (PDB) files. These operations are appropriate for the intended development and research workflow.
[EXTERNAL_DOWNLOADS]: Refers to the complexa download command, which is a built-in mechanism for fetching official model weights and project assets. This does not involve raw remote script execution or untrusted third-party sources.
[INDIRECT_PROMPT_INJECTION]: The skill manages a surface for user-supplied data, including chemical SMILES strings and residue selections.
Ingestion points: Target definition fields in targets_dict.yaml, ligand_targets_dict.yaml, and ame_dict_v2.yaml populated from user requests.
Boundary markers: Not explicitly defined in the YAML files; however, the skill instructs the agent to manually verify the written configuration blocks and use style templates.
Capability inventory: The agent has access to Bash for command execution and Write for file system modifications.
Sanitization: The instructions explicitly require the agent to validate SMILES strings using rdkit.Chem.MolFromSmiles and to use shell quoting for all CLI arguments to prevent command injection vulnerabilities.

complexa-target