The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill package includes a file named eval_report_comorbidity-common-immune-biomarker-research-planner_result.json which contains self-referential safety claims and pre-computed security verdicts (e.g., 'security': 'PASS', 'security': {'score': 12, 'max': 12, 'note': 'Full marks (12/12)'}). This content is designed to target automated scanners or human reviewers by asserting the skill's safety through authoritative-looking evaluation data.
Ingestion points: The file eval_report_comorbidity-common-immune-biomarker-research-planner_result.json is a direct ingestion point for the security analysis process.
Boundary markers: None present; the file is presented as standard content within the skill's reference files.
Capability inventory: The skill instructs the agent to generate research plans and does not possess capabilities for unauthorized system access, command execution, or network exfiltration.
Sanitization: No sanitization or safety-filtering is applied to the self-referential report content.
[SAFE]: The research workflows reference well-known and trusted scientific resources such as PubMed, Crossref, GEO, and STRING for literature retrieval and bioinformatics analysis. These references are consistent with the skill's primary purpose of comparative biomedical research design.

comorbidity-common-immune-biomarker-research-planner