The Agent Skills Directory

[UNVERIFIABLE_DEPENDENCIES]: The skill requires the installation of a package named langsmith-fetch via pip. This package is not an official LangChain library, and its existence and safety cannot be verified through standard official channels.
[METADATA_POISONING]: The skill provides a resource link to https://github.com/langchain-ai/langsmith-fetch, implying the tool is an official LangChain project. However, this repository does not exist under the official LangChain organization, which is a deceptive practice that could lead users to trust unverified code.
[PERSISTENCE]: In the troubleshooting section, the skill instructs users to append sensitive API keys to their shell configuration files (~/.bashrc or ~/.zshrc) using echo 'export LANGSMITH_API_KEY="your_key"' >> ~/.bashrc. This creates a persistent record of credentials in a plain-text configuration file.
[DATA_EXPOSURE]: The skill encourages users to verify their setup by running echo $LANGSMITH_API_KEY, which prints the sensitive API key to the terminal output where it becomes part of the agent's context and conversation history.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of LangSmith execution traces.
Ingestion points: Fetches trace data via langsmith-fetch traces --format json.
Boundary markers: None; the agent is instructed to analyze the raw JSON output.
Capability inventory: Includes shell command execution and file system operations (mkdir, grep).
Sanitization: No sanitization or filtering of the ingested trace content is mentioned before the agent analyzes it for 'root cause analysis' or 'suggested fixes'.
[COMMAND_EXECUTION]: The skill relies heavily on executing shell commands to interact with the LangSmith CLI, manage local files, and search through exported data using grep.

langsmith-fetch