The Agent Skills Directory

[DATA_EXFILTRATION]: The skill attempts to locate and read various workspace configuration files (e.g., .mcp.json, .claude/settings.json, .gemini/settings.json) to identify misconfigured arguments or missing environment variables. This file access is restricted to the primary troubleshooting purpose of the skill.
[EXTERNAL_DOWNLOADS]: Fetches troubleshooting guidelines and known issues from the official ChromeDevTools/chrome-devtools-mcp GitHub repository. This is documented neutrally as it targets a well-known service's official documentation.
[COMMAND_EXECUTION]: Executes diagnostic commands such as npx chrome-devtools-mcp@latest and gh issue list to verify the environment and search for existing bug reports. These tools are standard for the developer use-case intended by the skill.
[PROMPT_INJECTION]: The skill demonstrates an indirect prompt injection surface by ingesting external data from GitHub issues and documentation.
Ingestion points: Fetches content from Step 3 (troubleshooting documentation) and Step 6 (GitHub issue list).
Boundary markers: Absent; external content is processed without explicit delimiters or warnings to ignore embedded instructions.
Capability inventory: The agent has capabilities to execute shell commands (npx, gh) and read local workspace files.
Sanitization: No sanitization or validation of the remote GitHub content is performed before processing.

troubleshooting