The Agent Skills Directory

[OBFUSCATION]: The documentation file references/parsing_and_validation.md uses HTML entities (|) to represent the pipe character within markdown tables for regex patterns. While functionally intended to prevent markdown table breakage, this constitutes a form of encoding that masks raw logic.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from patient profiles, including disease types, molecular alterations, and treatment histories, which are then used to query external APIs and generate research reports.
Ingestion points: Patient clinical characteristics and molecular biomarkers provided by the user in SKILL.md.
Boundary markers: None detected; the skill does not instruct the agent to use specific delimiters when processing patient data for report generation.
Capability inventory: Significant network access via 40+ medical database tools and file system write capabilities for report synthesis.
Sanitization: Relies on regex parsing for identification but lacks explicit sanitization instructions for the final report output.
[DYNAMIC_EXECUTION]: The skill includes Python code snippets in references/parsing_and_validation.md and references/phases_detail.md that provide logic for parsing biomarkers and scoring matches, intended to be followed or executed by the agent.
[EXTERNAL_DOWNLOADS]: The skill fetches clinical and molecular data from established services including ClinicalTrials.gov, OpenTargets, CIViC, ChEMBL, and the FDA.

tooluniverse-clinical-trial-matching