The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied drug names and toxicological endpoints to generate research plans, creating a surface for indirect prompt injection. Additionally, the skill includes a documentation file (eval_report_single-drug-adverse-effect-hub-first-network-pharmacology_result.json) that contains self-referential safety claims and scores ('security: PASS', '12/12'), which is a pattern used to influence automated security analyzers.
Ingestion points: User input parameters for drug and adverse effect in SKILL.md; static self-referential safety assertions in the evaluation report JSON.
Boundary markers: Absent; there are no instructions for the agent to use delimiters or ignore embedded instructions within user-provided strings.
Capability inventory: None; the skill is composed entirely of markdown and JSON documentation and lacks any capability to execute shell commands, perform network operations, or modify the file system.
Sanitization: Absent; the skill does not provide mechanisms for validating or escaping external input strings.
[NO_CODE]: The skill consists exclusively of markdown reference files and a JSON evaluation report. No executable scripts (.py, .js, .sh), binary files, or configuration files that trigger code execution were identified in the skill package.

single-drug-adverse-effect-hub-first-network-pharmacology