skills/mukul975/anthropic-cybersecurity-skills/implementing-api-schema-validation-security/Gen Agent Trust Hub
implementing-api-schema-validation-security
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes a Python script (scripts/agent.py) that processes external OpenAPI specifications and JSON payloads. This creates an attack surface where maliciously crafted input files could potentially influence agent behavior or cause unintended side effects during processing.
- Ingestion points: The load_openapi_spec and validate_request_payload functions in scripts/agent.py accept file paths for processing external content.
- Boundary markers: The script does not implement explicit delimiters or instructions for the agent to ignore embedded prompts within the analyzed specifications.
- Capability inventory: The script performs file reading (open) and writes findings to a local JSON report.
- Sanitization: The script uses yaml.safe_load() for YAML parsing, which mitigates basic YAML-based attacks, but does not prevent logic-based injection within the specification content.
- [METADATA_POISONING]: There is a discrepancy between the author identified in the SKILL.md frontmatter ('mahipal') and the copyright holder in the LICENSE file ('mukul975'). While this does not directly impact the execution safety of the skill, inconsistent metadata regarding provenance is a form of deception.
Audit Metadata