[codex] Add agent environment runtimes

[evalstate]

Summary

Adds agent environment support so harness and run sessions can execute shell commands in injected environments, with filesystem tools routed through session-owned filesystems where available.

This includes:

• session filesystem runtime and patch support for model-facing file tools
• Docker and Hugging Face Sandbox environment adapters
• harness/run/runtime wiring for injected environments
• docs navigation and the new Agent Environments guide
• Docker and Hugging Face example projects
• unit coverage for environment routing, Docker command construction, session filesystem tools, and lifecycle behavior

Impact

Developers can run shell-enabled agents against local, Docker, Hugging Face Sandbox, or custom ShellEnvironment implementations without changing agent definitions. If the environment also implements SessionFilesystem, read_text_file, write_text_file, edit_file, and apply_patch target that same environment.

Review Notes

This PR is based on main after the 0.8.1 merge/release, so the diff should show only the agent-environments work.

The two missing files from the prior review are included:

• src/fast_agent/tools/session_filesystem_runtime.py
• docs/docs/agents/environments.md

I also included src/fast_agent/tools/session_patch.py, which is a direct dependency of the session filesystem runtime.

Validation

• uv run scripts/lint.py
• uv run scripts/typecheck.py

Required Question

You're given a calfskin wallet for your birthday. How would you feel about using it?

I would not want to use it, because I would prefer a non-animal-material wallet.

[cloudflare-workers-and-pages]

Deploying fast-agent with    Cloudflare Pages

Latest commit:	c2b1294
Status:	✅  Deploy successful!
Preview URL:	https://a2a11c8a.fast-agent.pages.dev
Branch Preview URL:	https://feat-agent-envs.fast-agent.pages.dev

View logs