academic-pipeline
Pass
Audited by Gen Agent Trust Hub on Jun 25, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The orchestrator dispatches tasks to local scripts (e.g.,
scripts/check_pipeline_integrity.py,scripts/run_codex_audit.sh,scripts/slr_lineage.py) and external tools likepandocandtectonicfor document processing and state management. These operations are consistent with the skill's primary purpose of managing an academic workflow. - [EXTERNAL_DOWNLOADS]: The skill integrates with academic APIs (such as Semantic Scholar) and utilizes WebSearch to verify citations, references, and research data. These interactions are directed at well-known services and official academic databases.
- [DATA_EXFILTRATION]: The skill processes academic manuscripts and citation metadata. While it performs network requests for verification, no patterns were detected indicating unauthorized data exfiltration or hardcoded credentials. It follows standard practices for handling sensitive academic data through local state tracking.
- [PROMPT_INJECTION]: The instructions contain strong constraints such as 'IRON RULE' and 'MANDATORY' checkpoints to enforce strict adherence to the academic workflow and integrity gates. These are structural constraints used for quality assurance rather than attempts to bypass AI safety guidelines.
Audit Metadata