The Agent Skills Directory

[SAFE]: The skill is entirely instructional and does not contain any executable scripts, binaries, or commands. It functions as a complex prompt template for generating medical research frameworks.
[PROMPT_INJECTION]: No instructions were found that attempt to bypass safety filters or override system prompts. The skill includes specific "Out-of-scope" instructions to redirect the agent away from clinical treatment recommendations, which acts as a safety guardrail.
[DATA_EXFILTRATION]: There are no patterns involving the access of sensitive local files or the transmission of data to external servers. The skill mentions literature retrieval but emphasizes using verified sources like PubMed and DOI links.
[OBFUSCATION]: The content is presented in clear markdown and JSON. No Base64, hex encoding, zero-width characters, or homoglyph attacks were detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided research themes (e.g., "Ferroptosis in lung cancer"). While this is an ingestion point for untrusted data, the skill has no dangerous capabilities (such as network operations or file system writes) that could be triggered maliciously. It only generates a text-based research plan.
[DYNAMIC_CONTEXT_INJECTION]: No instances of shell command placeholders (e.g., backticks with exclamation marks) were found in the skill definitions.
[REMOTE_CODE_EXECUTION]: The skill mentions various bioinformatics tools (e.g., limma, DESeq2, ConsensusClusterPlus) as part of a recommended workflow, but it does not attempt to install or execute these tools on the user's system.

pcd-immune-oncology-research-planner