markdown-a11y-assistant
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes standard system commands such as
findandgit diffduring its file discovery phase to identify relevant markdown documentation within the user's project directory. - [INDIRECT_PROMPT_INJECTION]: The skill identifies and processes untrusted markdown files, which presents a potential surface for indirect injection if those files contain instructions targeting the agent.
- Ingestion points: The skill ingests markdown (
*.md) files discovered via shell commands in Phase 1. - Boundary markers: The instructions include clear boundaries, such as an explicit directive to never modify content inside code blocks or YAML front matter.
- Capability inventory: The skill can read the file system, write report and CSV files to the current working directory, and delegate content modifications to a sub-agent tool.
- Sanitization: Remediation logic is constrained to specific accessibility domains (e.g., WCAG 2.2 criteria), and the process includes a mandatory human-in-the-loop 'Review Gate' before any file modifications are applied.
Audit Metadata