publish-html-report
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill generates and uploads reports to GitHub Gists using the
gh gist createcommand. Although these are categorized as 'secret' gists, they remain publicly accessible to anyone with the URL, potentially exposing internal architectural details or performance metrics. - [COMMAND_EXECUTION]: The workflow relies on the GitHub CLI (
gh) to automate the creation of gists and the retrieval of raw URLs via the GitHub API. - [EXTERNAL_DOWNLOADS]: The generated HTML scaffold fetches styling and visualization libraries from well-known content delivery networks, specifically Tailwind CSS via
cdn.tailwindcss.comand Mermaid viajsdelivr.net. - [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by ingesting codebase analysis and performance data for interpolation into an HTML scaffold without defined sanitization steps. 1. Ingestion points: Architecture exploration data, candidate identification results, and dependency graphs. 2. Boundary markers: No markers are present to distinguish between report templates and user-controlled data. 3. Capability inventory: Uses
ghCLI for gist publishing and public preview construction. 4. Sanitization: No explicit escaping or content validation is performed before data is embedded in the report. - [DYNAMIC_EXECUTION]: The Mermaid visualization library is initialized with
securityLevel: "loose". This configuration allows for more complex rendering but also enables the execution of HTML and scripts embedded within diagram definitions, which could lead to Cross-Site Scripting (XSS) if report content is malicious.
Audit Metadata