conference-tweet-generator
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external academic conference abstracts, which are untrusted data sources that could potentially contain malicious instructions intended to influence the agent's behavior.
- Ingestion points: The workflow in SKILL.md specifically involves ingesting and filtering abstracts provided by the user or found via external search.
- Boundary markers: The skill documentation includes a 'Validation and Safety Rules' section that explicitly instructs the agent not to fabricate information and to validate inputs before execution.
- Capability inventory: The skill utilizes the 'tavily_search' tool and executes a local validation script 'scripts/validate_skill.py'.
- Sanitization: The implementation uses a static reference 'references/drug_dictionary.md' to ensure controlled and accurate drug name translations, reducing the risk of injection-driven fabrication.
- [SAFE]: The script 'scripts/validate_skill.py' is a benign helper that performs simple argument checking with no access to sensitive system resources or network functions.
- [SAFE]: The audit metadata in 'conference-tweet-generator_audit_result_v2.json' is informational and does not contain executable code or patterns that impact the current security posture.
Audit Metadata