The Agent Skills Directory

Secure Data Parsing: The validate_catalog.py script implements a custom YAML loader derived from yaml.SafeLoader. This prevents the execution of arbitrary Python objects during catalog parsing, a common security consideration when handling YAML from external sources.
Path Traversal Mitigation: The file resolution logic in scripts/validate_catalog.py ensures that processed files are relative to permitted root directories and prevents the use of .. segments. This restricts file access to the intended catalog scope and prevents unauthorized read operations on the host system.
Input Validation and Schema Enforcement: The skill uses JSON Schema to enforce a strict structural contract for data catalogs. This ensures that only expected data types and fields are processed, reducing the risk of unexpected agent behavior from malformed or malicious inputs.
Resource Constraints: An operational limit of 5MB is enforced on catalog files before they are read into memory. This serves as a safeguard against resource exhaustion or denial-of-service attempts using extremely large input files.
Indirect Prompt Injection Considerations: The skill is designed to process external data inputs to generate catalogs, which creates a surface for potential instruction injection.
Ingestion points: Data enters the process via user-provided engagement context and inventory details as described in SKILL.md.
Boundary markers: The process is governed by the DS_CATALOG_V1 machine contract and success criteria that require schema validation.
Capability inventory: The skill uses a local script (scripts/validate_catalog.py) that performs controlled file system reads.
Sanitization: All input processing is mitigated by path sanitization and a secure, restricted YAML parser that rejects advanced features like anchors and aliases.

ds-catalog