The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill defines workflows that require the agent to ingest and analyze potentially untrusted content from user-provided Python source files and project directories (e.g., during auditing and refactoring tasks).
Ingestion points: Python source files, project configuration files (like pyproject.toml), and directory structures scanned during naming, type-checking, and audit workflows.
Boundary markers: The instructions include strong directives to avoid "performative agreement" and maintain technical rigor (found in references/code-review.md), which serve as conceptual guardrails against instructions that might be embedded in user code.
Capability inventory: The agent utilizes searching tools (ripgrep), linters (ruff), and is provided with Python templates capable of modifying the user's filesystem.
Sanitization: No explicit sanitization or escaping is defined for the content read from user files before it is processed by the agent's logic.
[DYNAMIC_EXECUTION]: The skill provides Python code templates in the "Bulk Operations" section of SKILL.md that allow the agent to generate and execute scripts for performing large-scale file transformations and metadata analysis.
Evidence: The templates in SKILL.md and examples/bulk_refactor.py use pathlib.Path and re to programmatically modify multiple files in the workspace.
[EXTERNAL_DOWNLOADS]: The skill workflows rely on the uv tool to manage and fetch Python dependencies from public registries based on project metadata or PEP 723 script blocks.
Evidence: Reference files like references/uv-projects.md and references/uv-scripts.md describe using uv add and uv run to resolve and install external packages.

python-ultimate