fineenvs ยท diff
v1.0 to v1.1.0
59 added, 82 removed. Audit A to A.
---
name: fineenvs
description: >
- Operate adithya-s-k/FineEnvs, a source-verified collection of RL environments, datasets,
- OpenEnv/ORS/NeMo Gym/Verifiers/SkyRL/GEM ports, reward evidence, and GRPO training recipes.
- Use when the user names FineEnvs, asks to run or inspect one of its environment projects, wants
- its deterministic graders or evaluation evidence, or wants to author a new environment from a
- description with the bundled framework skills. Route generic model fine-tuning, RL algorithm,
- or serving work to the appropriate generic skill instead.
+ Inspect and run the source-verified adithya-s-k/FineEnvs environment and dataset projects,
+ including OpenEnv, ORS, NeMo Gym, Verifiers, SkyRL Gym, and GEM dialects. Use when the user
+ names FineEnvs, supplies a checkout, asks for a local smoke or reward-evidence readout, or
+ wants to author an environment against its bundled contracts. Route generic training, RL
+ algorithms, serving, and provider operations to their owning skills.
allowed-tools: Bash Read Write Edit Glob Grep
compatibility: >
- Python repository using uv. Most examples require Python 3.11+; NeMo Gym requires Python 3.12+.
- HF Jobs, HF Hub, E2B, Daytona, OpenAI/HF inference, paid Spaces, CUDA GPUs, and model/data
- downloads are optional external services and must be explicitly approved before use.
+ Python repository using uv. Most projects require Python 3.11+; the optional NeMo Gym path
+ requires Python 3.12+. Cloud services, credentials, downloads, sandboxes, and GPUs are opt-in.
license: Apache-2.0
metadata:
- version: "1.0"
+ version: "1.1.0"
source: https://github.com/adithya-s-k/FineEnvs
upstream_commit: 5e0b46e96bfd20105976e50a06a3b79c9c62f076
---
# FineEnvs operator
- FineEnvs is an Apache-2.0 Python repository audited against upstream commit
- `5e0b46e96bfd20105976e50a06a3b79c9c62f076` (2026-10-12). Treat the pinned checkout and its
- project-local READMEs/configs as the source of truth. The repository contains runnable examples,
- not a guarantee that every historical training result can be reproduced: several recipes resolve
- moving upstream `main` dependencies, require paid services, or document known service/runtime gaps.
+ FineEnvs is an Apache-2.0 Python repository audited at
+ `5e0b46e96bfd20105976e50a06a3b79c9c62f076` (2026-10-12). Its project-local READMEs and configs
+ remain authoritative; historical runs, moving `main` dependencies, paid services, and known
+ runtime gaps are not reproducibility guarantees.
## When to use this skill
- - Inspect, pin, set up, smoke-test, or troubleshoot a FineEnvs checkout.
- - Run a no-backend Wordle rollout, an E2B-backed Jupyter/Desktop rollout, or an OpenEnv/ORS/NeMo Gym service.
- - Work with the LaTeX OCR, Watercolour, GeoGuesser, SmolDataEnvs, multi-harness RL, or PortSimEnv projects.
- - Read reward/assessment evidence before proposing training or comparing checkpoints.
- - Use the bundled `rl-env-from-description` and four framework-specific skills to author a new env.
+ - Inspect or pin a FineEnvs checkout and select one environment/project path.
+ - Run a no-backend Wordle smoke or another explicitly approved local fixture check.
+ - Read reward, grader, coverage, data-license, or framework-contract evidence.
+ - Author a new environment using the bundled domain and framework-generator contracts.
- Use the manifest-verified `soup` skill for model fine-tuning design unrelated to this checkout; use FineEnvs' bundled framework generator skills for standalone OpenEnv, ORS, Verifiers, or NeMo Gym work; use provider documentation or a separately installed provider skill for unrelated Hub or Jobs operations. External provider CLIs and services are not included in this skill.
+ Do not use this skill as a generic training, serving, Ray, sandbox, Hugging Face, provider, or
+ cloud-operations guide. Use the owning skill for those operations; consult the audited reference
+ here to identify their prerequisites and side effects.
## Instructions
- 1. **Pin and inspect before execution.** For a reproducible checkout, clone the repository and
- `git checkout 5e0b46e96bfd20105976e50a06a3b79c9c62f076`. Read the relevant project README and
- `pyproject.toml`, `README.md`, `REPRODUCE.md`, `VALIDATION.md`, or `eval/README.md` before
- choosing a command. Read `references/upstream.md` for the audited matrix and source links.
- 2. **Classify the operation.** Separate (a) read-only inventory, dry-run, local fixture smoke, and
- local CPU rollout from (b) service startup, external data/model downloads, credentials, HF Hub
- writes, paid inference, E2B/Daytona sandboxes, GPU Jobs, or deployed Spaces. Ask for explicit
- approval before category (b), and state the exact provider, model, hardware, data volume, and
- expected spend. Never print or commit secret values.
- 3. **Choose the smallest verified path.** Wordle is the cheapest end-to-end example. From its
- framework directory, `uv sync && uv run python rollout.py` uses the project-root `.env`; do not
- create per-folder secrets. Jupyter/Desktop require `E2B_API_KEY` for local/in-process use.
- `HF_TOKEN` drives HF inference/Jobs and `OPENAI_API_KEY` is only for OpenAI-native models.
- 4. **Preserve framework contracts.** OpenEnv is HTTP+MCP and discovers tools through
- `MCPToolClient.list_tools()`; ORS is HTTP REST+SSE and uses
- `EnvironmentsAPI(base_url=..., api_key="")` with per-tool `ToolOutput.reward`; NeMo Gym is
- HTTP with cookie sessions (`requests.Session`) and post-episode `/verify`; Verifiers, SkyRL Gym,
- and GEM are in-process. Do not replace discovery with guessed endpoints or silently treat a
- deployed-only NeMo Gym Space as locally runnable.
- 5. **Collect reward evidence, not just a score.** For SmolDataEnvs report grader mode, split,
- coverage, ungraded failures, and tool/token metrics with pass@1. For PortSimEnv report feasible,
- optimal, ended-without-submit, and reward distribution. For aesthetic or visual rewards report
- scorer availability and reward breakdown. A two-step smoke proves execution/checkpoint reload,
- not learning or quality improvement.
- 6. **Training is opt-in and bounded.** Prefer fixture/dry-run/evaluation first, then a small smoke
- with explicit output and budget. HF Jobs commands can create GPU, bucket, model, dashboard, and
- sandbox spend; pass secrets only through the provider's secret mechanism. Stop and report the
- exact blocker when prerequisites (GPU, Ray, E2B, Daytona, scorer Space, imagery, or inference
- quota) are absent; do not silently downgrade provider or hardware.
- 7. **Keep provenance with results.** Record the pinned FineEnvs commit, project path, resolved
- dependency commits/versions, model/checkpoint id, task split/identities, command, environment
- endpoint, hardware, reward/scorer settings, coverage, and output directory. Preserve Apache-2.0
- notices; PortSim data/viewer assets have additional CC BY-SA 4.0 and ODbL attribution requirements.
- 8. **For new environments**, interview for goal/action/state/reward/backend/termination/path, keep
- one framework-independent domain module, then invoke the bundled `rl-env-from-description` or a
- framework-specific generator. Finish with framework-specific import/server/tool-discovery/rollout
- checks; do not start training as part of authoring.
+ 1. **Preflight offline.** From the repo root run
+ `python3 .agent-skills/fineenvs/scripts/preflight.py --json`; from an installed skill root run
+ `python3 "$SKILLS_ROOT/fineenvs/scripts/preflight.py" --json`. For a checkout, add
+ `--repo PATH --expect-commit 5e0b46e96bfd20105976e50a06a3b79c9c62f076`. Treat
+ `runtime_verified:false` as inspection only: this helper does not import frameworks, read
+ secrets, query HTTP/GPU state, create sandboxes, or run evaluations. Optional framework
+ markers are informational; do not require all dialects at once.
+ 2. **Pin and read.** Verify the supplied checkout commit, then read its project-local README and
+ relevant `pyproject.toml`, `REPRODUCE.md`, `VALIDATION.md`, or `eval/README.md`. Read
+ [`references/upstream.md`](references/upstream.md) for the audited matrix, source links,
+ framework contracts, reward caveats, licensing, branch requirements, and cloud effects.
+ 3. **Classify before acting.** Keep inventory, source inspection, and no-backend local fixtures
+ separate from credentials, downloads, service startup, provider inference, E2B/Daytona or HF
+ sandboxes, paid Spaces, GPU Jobs, Hub/bucket writes, and deployment. Obtain explicit approval
+ for the latter and state provider, model, hardware, data volume, output path, and spend.
+ 4. **Choose the smallest path.** Wordle Verifiers is the low-cost local route:
+ `cd 00-environments-101/envs/wordle/verifiers && uv sync && uv run python rollout.py`.
+ Keep repository-root `.env` handling as documented; never invent per-project secret files.
+ 5. **Preserve evidence.** Report the pinned commit, project path, resolved versions or commits,
+ split, model/checkpoint, command, endpoint, hardware, reward/scorer settings, coverage, and
+ output directory. A smoke proves execution/checkpoint reload, not learning or quality.
+ 6. **Author safely.** Keep one framework-independent domain module, gather goal/action/state/
+ reward/backend/termination/path requirements, then use the bundled authoring or framework
+ generator skill. Finish with the selected framework's import, discovery, and rollout checks;
+ do not start training as part of authoring.
## Examples
```bash
- # Lowest-cost local smoke (no external backend)
+ # Offline host/check-out inspection; no upstream process is launched.
+ python3 .agent-skills/fineenvs/scripts/preflight.py --json
+ python3 .agent-skills/fineenvs/scripts/preflight.py --json \
+ --repo /path/to/FineEnvs --expect-commit 5e0b46e96bfd20105976e50a06a3b79c9c62f076
+
+ # Lowest-cost local smoke after reading the pinned project instructions.
cd 00-environments-101/envs/wordle/verifiers
uv sync && uv run python rollout.py
- # LaTeX OCR: offline fixture tests and package entrypoint
- uv run --project 01-latex-ocr/envs/latex_ocr --extra dev pytest 01-latex-ocr/envs/latex_ocr/tests -q
+ # Offline LaTeX OCR fixture path (the --real-data variant is not offline).
+ uv run --project 01-latex-ocr/envs/latex_ocr --extra dev \
+ pytest 01-latex-ocr/envs/latex_ocr/tests -q
uv run --project 01-latex-ocr/envs/latex_ocr latex-ocr-smoke
-
- # SmolDataEnvs: all rollout/reward paths use a Hugging Face Sandbox and external data/grader access
- cd 04-smoldataenvs
- # Requires explicit HF Sandbox approval, HF auth/token and quota, dependency/download approval, and output-path approval.
- uv run scripts/rollout.py
- uv run scripts/train_grpo.py --dry-run
- uv run scripts/eval_pass1.py --dry-run
- # `--dry-run` is not local: eval also downloads the dataset/grader and writes eval_results.json.
-
- # Multi-harness: bounded local smoke only after setup approval
- cd 05-multi-harness-rl
- python jobs/smoke.py --mode multi_harness --output runs/harbor-smoke
```
## Best practices
- - Read project-local instructions at the pinned revision; never assume the root quickstart covers
- every subproject.
- - Keep local services on localhost and use the documented endpoint variables (`OPENENV_URL`,
- `ORS_URL`, `NEMO_GYM_URL`, or project-specific equivalents); do not probe production Spaces.
- - Treat historical tables as evidence with their stated caveats, not as a current benchmark claim.
- - For an existing checkpoint, compare the same fixed split and report coverage alongside pass@1.
- - Do not use this skill to authorize cloud spend, data licensing, or credentials; those remain human
- decisions.
+ - Read project-local instructions at the pinned revision; the root quickstart is not a universal contract.
+ - Keep local endpoints on localhost and use documented variables; never probe production Spaces.
+ - Treat historical tables as qualified evidence, not current benchmark claims; report coverage with pass@1.
+ - Preserve Apache-2.0 notices and PortSim CC BY-SA 4.0/ODbL attribution obligations.
+ - Do not authorize cloud spend, data licensing, credentials, or external side effects on a user's behalf.
## References
- - [Audited source matrix and command contracts](references/upstream.md)
+ - [Audited matrix, framework/reward contracts, licensing, and side effects](references/upstream.md)
- [Pinned FineEnvs source](https://github.com/adithya-s-k/FineEnvs/tree/5e0b46e96bfd20105976e50a06a3b79c9c62f076)
- [Pinned root README](https://github.com/adithya-s-k/FineEnvs/blob/5e0b46e96bfd20105976e50a06a3b79c9c62f076/README.md)
- [Apache-2.0 license evidence](https://github.com/adithya-s-k/FineEnvs/blob/5e0b46e96bfd20105976e50a06a3b79c9c62f076/LICENSE)