The Agent Skills Directory

Indirect Prompt Injection: The skill ingests data from external sources, including local agent code, LangSmith traces, and dataset records. This presents a surface for potential instructions embedded in that data to influence agent behavior.
Ingestion points: The skill reads local agent source code (SKILL.md Step 3) and processes LangSmith trace and dataset structures (via scripts/inspect_trace.py and scripts/inspect_dataset.py).
Capability inventory: The skill allows for file reading/writing and shell command execution to run its bundled inspection scripts.
Boundary markers: No explicit delimiters or boundary markers are instructed for the data being analyzed to prevent the agent from following instructions contained within those files.
Sanitization: There are no explicit instructions for sanitizing or escaping the content retrieved from the LangSmith API before it is processed by the agent.
Command Execution: The skill instructs the agent to execute bundled Python scripts (inspect_trace.py and inspect_dataset.py) to analyze LangSmith infrastructure. These operations are performed using the user's preferred Python environment.
Dynamic Execution: A primary function of the skill is to have the agent generate and write new Python scripts for evaluator logic and experiment runners based on the analysis of the project's structure.
External Dependencies: The skill requires the langsmith Python package. This is a well-known library from the skill's author (langchain-ai) used for LLM application observability.

langsmith-code-eval