The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/setup.sh modifies system shell profile files (~/.bashrc and ~/.zshrc) to persist environment variables such as LANGSMITH_API_KEY. This represents a persistence mechanism that modifies shell configuration.
[COMMAND_EXECUTION]: The scripts/setup.sh script is vulnerable to command injection. It accepts user input for an API key and appends it directly to shell profile files without sanitization, which could allow a malicious actor to inject arbitrary commands that execute when a new shell is opened.
[REMOTE_CODE_EXECUTION]: The skill downloads and executes an installation script for the LangSmith CLI from the official repository of a well-known service (langchain-ai) using a shell pipe pattern.
[EXTERNAL_DOWNLOADS]: The skill automates the installation of several software packages (langsmith, openai, openevals) from public registries using pip, npm, and yarn within the scripts/setup.sh file.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted user requests and interpolates them into YAML structures for routing decisions.
Ingestion points: User prompts analyzed in SKILL.md.
Boundary markers: None present.
Capability inventory: The skill uses Bash and WebFetch tools and includes scripts (setup.sh, quickstart.py) capable of network operations and file system modification.
Sanitization: No input validation or sanitization is implemented for the processed data.

langsmith