The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run a local Node.js server and perform health checks using the command line.
Evidence: node scripts/server.mjs and node -e 'fetch(...)' commands documented in SKILL.md.
[DATA_EXPOSURE]: The skill indexes and serves files from the local filesystem, specifically targeting the ~/.codex/generated_images directory.
Evidence: The scripts/server.mjs file uses node:fs modules to read directory contents and stream image files to the local web interface.
Mitigation: The server implementation includes a safePath function that uses path.resolve and prefix checking to prevent directory traversal attacks, ensuring the server cannot serve files outside the intended gallery root.
[INDIRECT_PROMPT_INJECTION]: The web interface processes and displays metadata (such as filenames) originating from the local filesystem, which could theoretically contain malicious scripts if the filenames were attacker-controlled.
Evidence: Filenames and session IDs are rendered in assets/index.html.
Mitigation: The frontend includes an escapeHtml utility function that is consistently applied to image metadata before rendering, mitigating potential Cross-Site Scripting (XSS) vulnerabilities.

codex-image-gallery