The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains no instructions to bypass safety filters or override system behavior. It uses standard operational constraints to guide the research process and maintain a professional output format.\n- [DATA_EXFILTRATION]: No data exfiltration or sensitive file access patterns were detected. All network communication occurs through trusted academic tools directed at well-known scientific databases like PubMed, UniProt, and GTEx.\n- [REMOTE_CODE_EXECUTION]: The skill does not execute remote code, download unverified scripts, or use shell-based tools. It relies on a pre-defined set of research and annotation tools.\n- [OBFUSCATION]: No forms of obfuscation, such as base64 encoding, zero-width characters, or homoglyphs, were found in the instructions or reference files.\n- [INDIRECT_PROMPT_INJECTION]: While the skill ingests information from external scientific publications, the risk of malicious instruction injection is minimal due to the curated nature of the sources and the skill's structured synthesis process. This represents a low-severity inherent risk of data-processing skills.\n- [COMMAND_EXECUTION]: The skill performs no direct system command execution or arbitrary shell interactions. All operations are performed via high-level API calls to specialized databases.

tooluniverse-literature-deep-research