The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a Python utility script (scripts/ontology.py) to manage a local knowledge graph database. The script implements standard CRUD operations, relationship management, and schema validation via CLI arguments.
[DATA_EXPOSURE]: The scripts/ontology.py script includes a resolve_safe_path function that validates all file paths against the current working directory. This implementation prevents directory traversal attacks by ensuring the agent only reads or writes to the intended workspace root.
[CREDENTIALS_UNSAFE]: The ontology schema definitions for Account and Credential types include explicit constraints to prevent the storage of secrets. It mandates the use of secret_ref and specifically forbids properties like password, token, or api_key within the ontology graph, promoting secure secret management practices.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and store data from various sources into its graph structure. While this creates a surface for indirect prompt injection via processed data, the system includes a validation layer and the scripts treat all properties as static data, mitigating the risk of instructions being executed at the system level.
[DYNAMIC_EXECUTION]: The Python script uses yaml.safe_load to process schema definitions, which is a security best practice that prevents the execution of arbitrary code during YAML parsing.

ontology