acceptance-criteria-validate

Warn

Audited by Gen Agent Trust Hub on Mar 12, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is instructed to parse and execute shell commands located in the '## VALIDATION COMMANDS' section of implementation plan files. Although it includes instructions to avoid destructive operations, this mechanism allows for arbitrary command execution driven by file content.
  • Evidence: Step 2 'Run validation commands' instructs the agent to run commands found in the plan file's validation section.
  • [PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface by ingesting and acting upon instructions found in external, potentially untrusted markdown files.
  • Ingestion points: Reads plan files (e.g., '.agents/plans/*.md'), 'acceptance_criteria.md', and user request messages to determine validation logic.
  • Boundary markers: No specific delimiters or safety instructions are defined to separate ingested data from agent instructions.
  • Capability inventory: The agent has extensive capabilities including reading any file in the codebase, executing shell commands, and writing files to the '.agents/acceptance-validations/' directory.
  • Sanitization: No sanitization or validation of the content found in the plan files or criteria documents is performed before it is used to drive the agent's logic.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 12, 2026, 04:49 AM