The Agent Skills Directory

[SAFE]: The skill implements a set of Python scripts and documentation for performing statistical calculations required for pharmaceutical and clinical method validation. All calculations are performed using standard library functions, and no external dependencies or network operations are required.
[DYNAMIC_EXECUTION]: Several scripts (e.g., check_accuracy_precision.py, check_response.py, plan_validation.py) use the __import__ function. Analysis confirms this is used for standard path resolution to locate and import local utility modules (_common.py and _catalog.py) relative to the script's location. This is a benign programming pattern for standalone scripts and does not load untrusted external code.
[DATA_EXPOSURE_AND_EXFILTRATION]: The scripts read user-supplied CSV, TSV, or JSON data for statistical processing. The analysis found no code that accesses sensitive system paths, hardcodes credentials, or attempts to transmit data over the network.
[COMMAND_EXECUTION]: The skill requires Bash access to execute the provided Python scripts. The commands demonstrated in the SKILL.md and documentation are restricted to running these local scripts with parameters. No high-risk command execution (e.g., privilege escalation, persistence) was found.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from CSV/JSON files. However, it uses safe parsers (csv.DictReader and json.loads) and limits the scope of operations to mathematical analysis. The scripts also include defensive checks, such as MAX_INPUT_BYTES and MAX_ROWS, to mitigate resource exhaustion attacks.
[PROMPT_INJECTION]: The instructions include safety-oriented guidelines, such as advising the agent not to attempt to retrieve or reconstruct paywalled/copyrighted standard text (USP, CLSI), which serves as a protective constraint rather than a malicious bypass.

analytical-method-validation