dbs-wechat-html
Pass
Audited by Gen Agent Trust Hub on Oct 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill performs local document conversion and formatting tasks without initiating network connections or accessing sensitive system files.
- [COMMAND_EXECUTION]: The agent is instructed to use standard command-line utilities (
rgfor ripgrep andxmllint) to perform a 'static check' on the generated HTML files. These commands verify the absence of forbidden elements like<script>tags or external CSS which are incompatible with the target platform. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided Markdown content which could theoretically contain malicious instructions.
- Ingestion points: Markdown files or raw text input supplied by the user (SKILL.md).
- Boundary markers: The skill defines a narrow task scope, explicitly stating that it should not rewrite views or diagnose content, only perform layout tasks.
- Capability inventory: The agent can write files to a specific output directory and execute validation commands (
xmllint,rg). - Sanitization: The skill mandates the escaping of HTML special characters to prevent structural breakage.
- [DATA_EXPOSURE_AND_EXFILTRATION]: The instructions explicitly forbid the use of external URLs, scripts, or network-loaded resources, ensuring that processed data remains within the local environment.
Audit Metadata