The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process user-provided RDF and OWL files, which represents a surface for untrusted data to enter the agent's context.
Ingestion points: External data enters via .rdf and .owl files provided by the user, as described in SKILL.md.
Boundary markers: The instructions do not explicitly define delimiters or specific warnings to ignore instructions that might be embedded within the ontology file content.
Capability inventory: The skill has the capability to write to the repository's filesystem (specifically under the catalogue/ directory) and execute local validation commands.
Sanitization: While the skill utilizes a validation script (scripts/compile-catalogue.ts), it does not specify internal sanitization or escaping mechanisms for the raw input data within the instructions.
Local Command Execution: The skill automates the execution of local scripts and build tools to ensure catalogue entries are valid.
Evidence: The 'Validate' section in SKILL.md instructs the agent to run npx tsx scripts/compile-catalogue.ts and npm run validate. This allows the agent to interact with the local shell to perform necessary compilation and integrity checks on the imported data.

ontology-catalog-import