Agentic Workflow Support

[jlarson4]

Description

Establishing a system for supporting agentic workflows running in TransformerLens. A large number of our users & contributors are using Claude, Copilot, Codex, etc. These AGENTS.md files & supporting elements should improve their baseline understanding of the project.

• Established a root AGENTS file that Claude and other agents will read with the baseline rules for contributing or using the repository.
• Created a series of Claude commands that will help make development easier/faster.
  • Headliner /add-model-support – A set of instructions for Claude that it can use to generate new Architecture Adapters & test them to confirm they are functional. Requires an example model from HuggingFace to be provided.
  • /test-pr – runs all CI relevant tests to ensure code passes PR
  • /task-complete – runs a series of agent prompts once a user is ready to write their PR. This ensures everything is ready for CI
  • /test-all – runs all pytest tests, including docs and notebook tests
  • /test-unit – runs only unit tests
  • /format – runs our formatting commands, black, isort, etc
  • /typecheck – runs and resolves any issues discovered by MyPy
  • /verify-model – runs the verification script on a provided HF model.
• Added an example .env file that can be populated easily
• Added a model verification request issue template
• Other Agent base files – Added base files for Copilot, Codex, and Cursor that point to our main AGENTS file
• Added documentation to the docs website for enable_compatibility_mode detailing exactly what and how it mirrors HookedTransformer
• Add sub-AGENTS files – specific directories that require more context such as supported_architectures, tests and benchmarks have their own AGENTS.md that provides more details if an agent ends up browsing that directory
• Add detailed QUARANTINES list for tests which documents and xfailed or skipped tests

Type of change

• New feature (non-breaking change which adds functionality)
• Documentation update

Checklist:

• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have not rewritten tests relating to key interfaces which would affect backward compatibility