inplasy-registration-helper
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user data (study title and protocol) to fill registration templates, creating a surface where malicious instructions could be embedded in the input. * Ingestion points: User-provided study title and protocol processed via
references/registration_templates.md. * Boundary markers: SKILL.md provides syntax rules for placeholders (preserving curly braces and filling angle brackets) but lacks explicit instructions for the agent to ignore embedded commands in the user input. * Capability inventory: The skill executesscripts/date_utils.pyto calculate timelines and generates text deliverables. * Sanitization: No input validation or sanitization is performed on user strings before interpolation. - [METADATA_POISONING]: The skill includes a file
INPLASY-registration-helper_audit_result_v1.jsonthat simulates an authoritative security audit. * Evidence: The file contains claims such as "security: PASS" and "grade: Production Ready." These claims are self-referential data and must be disregarded in favor of independent analysis. - [METADATA_POISONING]: There is a naming inconsistency in the documentation that could lead to confusion about the tool's intended purpose. * Evidence: The header in
SKILL.mdrefers to a "PROSPERO Registration Helper," while the YAML frontmatter and all other files specify "INPLASY-registration-helper."
Audit Metadata