The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Fetches documentation, integration guides, and framework-specific instructions from the vendor's official domain (langwatch.ai) and GitHub repositories.
[REMOTE_CODE_EXECUTION]: Utilizes npx to download and execute the langwatch report diagnostic tool from the npm registry. This is used for reporting errors to the vendor and is gated by explicit user approval.
[COMMAND_EXECUTION]: Executes shell commands to install the LangWatch SDK using various package managers (pip, npm, uv, pnpm, yarn, bun). It also runs short scripts via the Python and Node.js interpreters to verify that environment variables like LANGWATCH_API_KEY are correctly configured.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read the full codebase and git history to determine the scope for tracing. This introduces a surface where malicious instructions embedded in the project's source code or commit history could potentially influence the agent's behavior. However, the skill includes specific instructions to ignore non-essential secrets in environment files.
[DATA_EXFILTRATION]: Provides a mechanism to send diagnostic reports and session transcripts to the vendor. The skill explicitly requires a --user-approved flag and includes local scrubbing logic to remove PII and secrets before data leaves the user's environment.

tracing