The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructions direct the user to display sensitive API keys in the terminal output, which can be captured in shell history or logging systems.
Evidence: The 'Verify setup' section uses echo $LANGSMITH_API_KEY.
[COMMAND_EXECUTION]: The skill implements a persistence mechanism by instructing the agent to modify the user's shell profile (~/.bashrc) to permanently store sensitive credentials.
Evidence: The Troubleshooting section provides the command echo 'export LANGSMITH_API_KEY="your_key"' >> ~/.bashrc.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of the langsmith-fetch Python package from an external registry (PyPI), which is a third-party dependency not included in the platform's trusted libraries.
Evidence: pip install langsmith-fetch in the Prerequisites section.
[COMMAND_EXECUTION]: The skill workflows involve executing shell commands with user-provided arguments, such as <trace-id>, which are not explicitly sanitized, creating a potential surface for command injection.
Evidence: langsmith-fetch trace <trace-id> --format json in Workflow 2.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests and processes external JSON trace data from LangSmith without using boundary markers or sanitization to prevent the agent from following instructions embedded in the traces.
Ingestion points: langsmith-fetch trace <trace-id> --format json in SKILL.md.
Boundary markers: Absent.
Capability inventory: Shell command execution and file system modification (writing to ~/.bashrc).
Sanitization: Absent.

langsmith-fetch