The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it is designed to analyze and run content from untrusted external Jupyter notebook files (.ipynb).
Ingestion points: The skill ingests data from notebook cells (code and markdown) to diagnose and fix errors, as described in SKILL.md.
Boundary markers: Absent. There are no explicit delimiters or instructions provided to the agent to distinguish between its own logic and instructions that might be embedded in the notebook data.
Capability inventory: The agent has the ability to execute code in cells, run system commands via subprocess.run, and perform package management with pip and micromamba.
Sanitization: Absent. No input validation or sanitization of the notebook content is performed before processing or execution.
[COMMAND_EXECUTION]: The skill uses command-line tools for environment diagnostics and configuration.
Evidence: Instructions in SKILL.md and references/environment-management.md utilize pip show, pip freeze, and micromamba env export to inspect system state.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading and installing Python packages from external registries to resolve environment issues.
Evidence: Documentation includes examples of pip install, micromamba install, and pip install -r requirements.txt to fetch dependencies as part of the troubleshooting workflow.

notebook-debugger