write-validation

Pass

Audited by Gen Agent Trust Hub on Nov 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill documents the prompt method, which allows for LLM-based semantic validation of data. This creates a surface for indirect prompt injection if the data being validated contains malicious instructions. 1. Ingestion points: Data tables passed to pb.Validate(data=df) in SKILL.md. 2. Boundary markers: The documentation does not specify the use of delimiters or instructions to ignore embedded commands. 3. Capability inventory: Validation failures can trigger Actions such as Slack notifications or custom Python functions. 4. Sanitization: No sanitization methods for input data are described for the prompt check.
  • [COMMAND_EXECUTION]: The skill documentation includes examples of using Actions and FinalActions to execute Python callables (e.g., sys.exit(1), pb.send_slack_notification) in response to validation results, which is a standard feature for automated alerting and pipeline orchestration.
  • [EXTERNAL_DOWNLOADS]: The skill requires the pointblank Python package, as specified in the compatibility metadata in SKILL.md.
  • [DYNAMIC_EXECUTION]: The skill supports dynamic logic through the pre argument (e.g., lambda functions for data filtering) and the specially method, which allows users to define and execute custom validation logic at runtime.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 15, 2026, 07:31 AM