The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is instructed to search for and read various configuration files, including .mcp.json, .claude/settings.json, and .gemini/settings.json. These files frequently contain sensitive information such as API keys, environment variables, and authentication tokens for other MCP services. Accessing these files exposes potential secrets to the agent context.
[COMMAND_EXECUTION]: The skill executes shell commands for diagnostics, specifically npx chrome-devtools-mcp@latest and the GitHub CLI (gh). While these are used for troubleshooting purposes, they represent active command execution on the user's system.
[EXTERNAL_DOWNLOADS]: The skill fetches content from external sources, including documentation from the ChromeDevTools GitHub repository and potentially downloading the latest version of the chrome-devtools-mcp package via npx. These resources originate from the vendor's official repositories.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it reads troubleshooting guides and searches GitHub issues. Content from these external, attacker-influenced sources could contain malicious instructions intended to manipulate the agent's behavior during a troubleshooting session.
Ingestion points: Reads https://github.com/ChromeDevTools/chrome-devtools-mcp/blob/main/docs/troubleshooting.md and searches GitHub issues using the gh tool.
Boundary markers: None identified; external content is processed directly in the diagnostic flow.
Capability inventory: Includes local file reading and shell command execution (npx, gh).
Sanitization: No explicit sanitization or filtering of external GitHub content is performed before processing.

troubleshooting