loopgym 0.1.0__tar.gz

loopgym-0.1.0/.github/workflows/publish.yml

@@ -0,0 +1,33 @@
+name: Publish to PyPI
+
+on:
+  release:
+    types: [published]
+  workflow_dispatch:
+
+permissions:
+  id-token: write
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install build tools
+        run: pip install hatchling build
+
+      - name: Build package
+        run: python -m build
+
+      # Uses PYPI_API_TOKEN when set; otherwise OIDC trusted publishing.
+      - name: Publish to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          password: ${{ secrets.PYPI_API_TOKEN }}
+          skip-existing: true
+

loopgym-0.1.0/.github/workflows/test.yml

@@ -0,0 +1,39 @@
+name: test-loopgym
+
+on:
+  push:
+    branches: [main, master]
+  pull_request:
+    branches: [main, master]
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+
+      - uses: actions/checkout@v4
+        with:
+          repository: KanakMalpani/Loop-Core-Engineering
+          path: deps/loop-core
+
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+
+      - name: Install loopgym
+        run: pip install -e ".[dev]"
+
+      - name: Validate bundled LSS specs (Loop Core Engineering)
+        run: |
+          pip install jsonschema pyyaml
+          CORE="deps/loop-core"
+          for spec in envs/loopbench/*/spec.yaml examples/minimal-loop.yaml; do
+            python "$CORE/tools/validate_lss.py" "$spec" --schema "$CORE/specs/lss-1.0.schema.json"
+          done
+
+      - name: Run tests
+        run: pytest tests/ -q
+
+      - name: Quickstart smoke test
+        run: python examples/quickstart.py

loopgym-0.1.0/.gitignore

@@ -0,0 +1,16 @@
+__pycache__/
+*.py[cod]
+.venv/
+venv/
+.pytest_cache/
+.mypy_cache/
+.ruff_cache/
+*.egg-info/
+dist/
+build/
+.DS_Store
+.env
+.env.*
+!.env.example
+*.pem
+credentials.json

loopgym-0.1.0/CONTRIBUTING.md

@@ -0,0 +1,28 @@
+# Contributing to LoopGym
+
+## What belongs here
+
+- Environment implementations (SimEnv, LiveEnv, ReplayEnv)
+- LSS compiler / loop runtime
+- Env registry and bundled LoopBench env fixtures
+- Evaluators and CLI
+
+## What does not belong here
+
+- LSS schema changes — [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering)
+- Benchmark task definitions / LES scoring — [LoopBench](https://github.com/KanakMalpani/LoopBench)
+- Dataset records — [LoopNet](https://github.com/KanakMalpani/loopnet)
+
+## Before opening a PR
+
+```bash
+pip install -e ".[dev]"
+pytest tests/ -q
+python examples/quickstart.py
+```
+
+CI validates bundled LSS specs against [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering).
+
+## License
+
+MIT — see [LICENSE](LICENSE).

loopgym-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 KanakMalpani
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

loopgym-0.1.0/PKG-INFO

@@ -0,0 +1,188 @@
+Metadata-Version: 2.4
+Name: loopgym
+Version: 0.1.0
+Summary: OpenAI Gym equivalent for loops — create, run, benchmark, compare, evolve
+Project-URL: Homepage, https://github.com/KanakMalpani/LoopGym
+Project-URL: Repository, https://github.com/KanakMalpani/LoopGym
+Project-URL: Issues, https://github.com/KanakMalpani/LoopGym/issues
+Project-URL: Documentation, https://github.com/KanakMalpani/LoopGym/blob/main/docs/api.md
+Project-URL: Loop Core Engineering, https://github.com/KanakMalpani/Loop-Core-Engineering
+Author: Kanak Malpani
+License: MIT
+License-File: LICENSE
+Keywords: agents,loop-engineering,loopgym,reinforcement-learning
+Classifier: Development Status :: 4 - Beta
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3.12
+Requires-Python: >=3.12
+Requires-Dist: pyyaml>=6.0
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.4; extra == 'dev'
+Description-Content-Type: text/markdown
+
+<p align="center">
+  <strong>LoopGym</strong><br>
+  <em>OpenAI Gym for self-improving loops.</em>
+</p>
+
+<p align="center">
+  <a href="https://github.com/KanakMalpani/LoopGym/actions/workflows/test.yml"><img src="https://github.com/KanakMalpani/LoopGym/actions/workflows/test.yml/badge.svg" alt="CI"></a>
+  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="MIT"></a>
+  <img src="https://img.shields.io/badge/python-3.12+-blue.svg" alt="Python 3.12+">
+  <a href="https://github.com/KanakMalpani/Loop-Core-Engineering"><img src="https://img.shields.io/badge/LSS-1.0.0-green.svg" alt="LSS 1.0"></a>
+  <img src="https://img.shields.io/badge/envs-5-blue.svg" alt="5 environments">
+</p>
+
+---
+
+If **LSS** is how you *declare* a loop, **LoopGym** is how you *run* it.
+
+LoopGym compiles [LSS 1.0](https://github.com/KanakMalpani/Loop-Core-Engineering) YAML into executable environments — with deterministic simulation for CI, live model backends for production eval, and trajectory replay from [LoopNet](https://github.com/KanakMalpani/loopnet). One API. Three backends. Zero vendor lock-in on the spec.
+
+```python
+import loopgym as lg
+
+env = lg.make("loopbench/code-repair-v1")
+obs = env.reset(task_id="cr-001")
+while not env.done:
+    obs, reward, done, info = env.step(agent.action(obs))
+```
+
+<p align="center">
+  <a href="#-install--run"><strong>Install & run →</strong></a> ·
+  <a href="docs/api.md">API reference</a> ·
+  <a href="examples/quickstart.py">Quickstart script</a>
+</p>
+
+---
+
+## Why LoopGym
+
+| Problem | LoopGym answer |
+|---------|----------------|
+| Every benchmark rolls its own runner | Shared `loopgym.make(env_id)` registry |
+| CI can't afford API keys | **SimEnv** — deterministic, free, fast |
+| Production eval needs real models | **LiveEnv** — pluggable backends |
+| Historical analysis burns budget | **ReplayEnv** — LoopNet trajectories, no LLM calls |
+
+[LoopBench](https://github.com/KanakMalpani/LoopBench) defines tasks and scores them; LoopGym executes. Clean separation, like Gym vs. benchmark suites in RL.
+
+---
+
+## Architecture
+
+```mermaid
+flowchart TB
+  LSS[LSS YAML spec]
+  COMP[LoopGym compiler]
+  SIM[SimEnv]
+  LIVE[LiveEnv]
+  REPLAY[ReplayEnv]
+  BENCH[LoopBench runner]
+
+  LSS --> COMP
+  COMP --> SIM
+  COMP --> LIVE
+  COMP --> REPLAY
+  SIM --> BENCH
+  LIVE --> BENCH
+  REPLAY --> BENCH
+```
+
+---
+
+## ⚡ Install & run
+
+**One-liner (GitHub):**
+
+```bash
+pip install git+https://github.com/KanakMalpani/LoopGym.git
+python -c "import loopgym as lg; env = lg.make('loopbench/code-repair-v1'); print(env.reset(task_id='cr-001'))"
+```
+
+**Developer setup:**
+
+```bash
+git clone https://github.com/KanakMalpani/LoopGym.git && cd LoopGym
+pip install -e ".[dev]"
+python examples/quickstart.py
+pytest tests/ -q
+```
+
+**PyPI** (after first release — [PUBLISHING.md](PUBLISHING.md)):
+
+```bash
+pip install loopgym
+```
+
+---
+
+## Environments
+
+| Env ID | Backend | Use case |
+|--------|---------|----------|
+| `loopbench/code-repair-v1` | SimEnv | Verify-driven code repair |
+| `loopbench/research-synthesis-v1` | SimEnv | Research brief synthesis |
+| `loopbench/multi-agent-debate-v1` | SimEnv | Multi-agent review / debate |
+| `replay/loopnet-v1` | ReplayEnv | Replay [LoopNet](https://github.com/KanakMalpani/loopnet) trajectories |
+| `sim/mock-llm-v1` | SimEnv | Generic mock-LLM sandbox |
+
+Bundled LSS specs live under [`envs/loopbench/`](envs/loopbench/). All validated against [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering) in CI.
+
+---
+
+## LoopNet replay (optional)
+
+```bash
+git clone https://github.com/KanakMalpani/loopnet.git ../loopnet
+# or: export LOOPNET_SEED_PATH=/path/to/records.jsonl
+```
+
+```python
+env = lg.make("replay/loopnet-v1")
+env.reset(record_id="ln-00042")
+```
+
+---
+
+## Ecosystem
+
+| Repository | Role |
+|------------|------|
+| [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering) | LSS / LES authority |
+| [LoopNet](https://github.com/KanakMalpani/loopnet) | Trajectory corpus |
+| **LoopGym** | Runtime (this repo) |
+| [LoopBench](https://github.com/KanakMalpani/LoopBench) | Benchmark orchestration |
+
+Full stack map: [ECOSYSTEM.md](https://github.com/KanakMalpani/Loop-Core-Engineering/blob/main/ECOSYSTEM.md)
+
+---
+
+## Project layout
+
+| Path | Purpose |
+|------|---------|
+| [`loopgym/`](loopgym/) | Registry, envs, runtime, evaluators |
+| [`envs/loopbench/`](envs/loopbench/) | Task fixtures + LSS specs |
+| [`docs/api.md`](docs/api.md) | API reference |
+| [`examples/quickstart.py`](examples/quickstart.py) | Onboarding smoke test |
+
+---
+
+## Citation
+
+```bibtex
+@software{loopgym2026,
+  title={LoopGym: OpenAI Gym for LSS-Defined Agent Loops},
+  author={Malpani, Kanak},
+  year={2026},
+  url={https://github.com/KanakMalpani/LoopGym}
+}
+```
+
+---
+
+<p align="center">
+  <sub>MIT · v0.1 · <a href="CONTRIBUTING.md">Contributing</a> · <a href="SECURITY.md">Security</a> · <a href="PUBLISHING.md">PyPI</a></sub>
+</p>

loopgym-0.1.0/PLAN.md

@@ -0,0 +1,62 @@
+# 05 — loopgym
+
+## One-line purpose
+
+**LoopGym** — OpenAI Gym equivalent for loops: create, run, benchmark, compare, evolve, visualize.
+
+## Why this repo exists
+
+Adoption follows **runnable code**. LoopGym is the reference runtime that compiles LSS → executable loop environments.
+
+## Scope (in scope)
+
+- `loopgym.make(env_id)` API
+- Env backends:
+  - **SimEnv** — mock LLM + mock oracles (no API keys)
+  - **ReplayEnv** — replay LoopNet trajectories
+  - **LiveEnv** — real LLM APIs (optional, user keys)
+- LSS compiler → runtime graph
+- Evaluator plugins (command, rubric, deterministic)
+- Evolution module (optional v0.2): mutate LSS hyperparams
+- Integration hooks for LangGraph, generic Python
+
+## Scope (out of scope)
+
+- Official leaderboard hosting → `06-loopbench`
+- Dataset curation → `04-loopnet`
+
+## Deliverables v0.1
+
+- [x] `loopgym/` Python package
+- [x] `envs/loopbench/` stub envs (3 tasks)
+- [x] `examples/quickstart.py`
+- [x] `docs/api.md`
+- [ ] PyPI publish workflow (optional)
+
+## API sketch
+
+```python
+import loopgym as lg
+
+env = lg.make("loopbench/code-repair-v1", spec_path="my-loop.yaml")
+obs = env.reset(task_id="cr-001")
+while not env.done:
+    obs, reward, done, info = env.step(agent.action(obs))
+```
+
+## Dependencies
+
+- **01-loop-engineering-core** — LSS schema
+- **04-loopnet** — ReplayEnv data (optional v0.1)
+
+## Success criteria
+
+`quickstart.py` runs with MockLLM; same LSS runs on 3 seeds with reproducible trajectories in SimEnv.
+
+## Agent instructions
+
+Extract and generalize `implementations/generic/loop_runtime.py` from `Loop Engineering` repo; don't fork forever — migrate to this repo as canonical runtime.
+
+## Status
+
+✅ v0.1 shipped (2026-06-13)

loopgym-0.1.0/PUBLISHING.md

@@ -0,0 +1,43 @@
+# Publishing LoopGym to PyPI
+
+## One-time setup
+
+1. Create a project at [pypi.org](https://pypi.org/) named **`loopgym`** (register the name before first publish).
+2. **Preferred:** configure [trusted publishing](https://docs.pypi.org/trusted-publishers/) on the PyPI project:
+   - **PyPI project name:** `loopgym`
+   - **Owner:** `KanakMalpani`
+   - **Repository name:** `LoopGym`
+   - **Workflow name:** `publish.yml`
+   - **Environment name:** *(leave blank)*
+
+   Linking GitHub under your PyPI account is not enough — each project needs its own trusted publisher entry.
+
+3. **Fallback:** add **`PYPI_API_TOKEN`** (upload scope) to Settings → Secrets → Actions on this repo.
+
+## Publish
+
+Create a GitHub Release (tag `v0.1.0` → publishes `0.1.0`):
+
+```bash
+git tag v0.1.0
+git push origin v0.1.0
+gh release create v0.1.0 --title "v0.1.0" --notes "Initial public release"
+```
+
+Or run **Actions → Publish to PyPI → Run workflow** manually.
+
+## Install
+
+```bash
+pip install loopgym
+```
+
+## Verify locally before release
+
+```bash
+pip install build
+python -m build
+pip install dist/loopgym-*.whl
+loopgym --help
+pytest tests/ -q
+```

loopgym-0.1.0/README.md

@@ -0,0 +1,165 @@
+<p align="center">
+  <strong>LoopGym</strong><br>
+  <em>OpenAI Gym for self-improving loops.</em>
+</p>
+
+<p align="center">
+  <a href="https://github.com/KanakMalpani/LoopGym/actions/workflows/test.yml"><img src="https://github.com/KanakMalpani/LoopGym/actions/workflows/test.yml/badge.svg" alt="CI"></a>
+  <a href="LICENSE"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="MIT"></a>
+  <img src="https://img.shields.io/badge/python-3.12+-blue.svg" alt="Python 3.12+">
+  <a href="https://github.com/KanakMalpani/Loop-Core-Engineering"><img src="https://img.shields.io/badge/LSS-1.0.0-green.svg" alt="LSS 1.0"></a>
+  <img src="https://img.shields.io/badge/envs-5-blue.svg" alt="5 environments">
+</p>
+
+---
+
+If **LSS** is how you *declare* a loop, **LoopGym** is how you *run* it.
+
+LoopGym compiles [LSS 1.0](https://github.com/KanakMalpani/Loop-Core-Engineering) YAML into executable environments — with deterministic simulation for CI, live model backends for production eval, and trajectory replay from [LoopNet](https://github.com/KanakMalpani/loopnet). One API. Three backends. Zero vendor lock-in on the spec.
+
+```python
+import loopgym as lg
+
+env = lg.make("loopbench/code-repair-v1")
+obs = env.reset(task_id="cr-001")
+while not env.done:
+    obs, reward, done, info = env.step(agent.action(obs))
+```
+
+<p align="center">
+  <a href="#-install--run"><strong>Install & run →</strong></a> ·
+  <a href="docs/api.md">API reference</a> ·
+  <a href="examples/quickstart.py">Quickstart script</a>
+</p>
+
+---
+
+## Why LoopGym
+
+| Problem | LoopGym answer |
+|---------|----------------|
+| Every benchmark rolls its own runner | Shared `loopgym.make(env_id)` registry |
+| CI can't afford API keys | **SimEnv** — deterministic, free, fast |
+| Production eval needs real models | **LiveEnv** — pluggable backends |
+| Historical analysis burns budget | **ReplayEnv** — LoopNet trajectories, no LLM calls |
+
+[LoopBench](https://github.com/KanakMalpani/LoopBench) defines tasks and scores them; LoopGym executes. Clean separation, like Gym vs. benchmark suites in RL.
+
+---
+
+## Architecture
+
+```mermaid
+flowchart TB
+  LSS[LSS YAML spec]
+  COMP[LoopGym compiler]
+  SIM[SimEnv]
+  LIVE[LiveEnv]
+  REPLAY[ReplayEnv]
+  BENCH[LoopBench runner]
+
+  LSS --> COMP
+  COMP --> SIM
+  COMP --> LIVE
+  COMP --> REPLAY
+  SIM --> BENCH
+  LIVE --> BENCH
+  REPLAY --> BENCH
+```
+
+---
+
+## ⚡ Install & run
+
+**One-liner (GitHub):**
+
+```bash
+pip install git+https://github.com/KanakMalpani/LoopGym.git
+python -c "import loopgym as lg; env = lg.make('loopbench/code-repair-v1'); print(env.reset(task_id='cr-001'))"
+```
+
+**Developer setup:**
+
+```bash
+git clone https://github.com/KanakMalpani/LoopGym.git && cd LoopGym
+pip install -e ".[dev]"
+python examples/quickstart.py
+pytest tests/ -q
+```
+
+**PyPI** (after first release — [PUBLISHING.md](PUBLISHING.md)):
+
+```bash
+pip install loopgym
+```
+
+---
+
+## Environments
+
+| Env ID | Backend | Use case |
+|--------|---------|----------|
+| `loopbench/code-repair-v1` | SimEnv | Verify-driven code repair |
+| `loopbench/research-synthesis-v1` | SimEnv | Research brief synthesis |
+| `loopbench/multi-agent-debate-v1` | SimEnv | Multi-agent review / debate |
+| `replay/loopnet-v1` | ReplayEnv | Replay [LoopNet](https://github.com/KanakMalpani/loopnet) trajectories |
+| `sim/mock-llm-v1` | SimEnv | Generic mock-LLM sandbox |
+
+Bundled LSS specs live under [`envs/loopbench/`](envs/loopbench/). All validated against [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering) in CI.
+
+---
+
+## LoopNet replay (optional)
+
+```bash
+git clone https://github.com/KanakMalpani/loopnet.git ../loopnet
+# or: export LOOPNET_SEED_PATH=/path/to/records.jsonl
+```
+
+```python
+env = lg.make("replay/loopnet-v1")
+env.reset(record_id="ln-00042")
+```
+
+---
+
+## Ecosystem
+
+| Repository | Role |
+|------------|------|
+| [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering) | LSS / LES authority |
+| [LoopNet](https://github.com/KanakMalpani/loopnet) | Trajectory corpus |
+| **LoopGym** | Runtime (this repo) |
+| [LoopBench](https://github.com/KanakMalpani/LoopBench) | Benchmark orchestration |
+
+Full stack map: [ECOSYSTEM.md](https://github.com/KanakMalpani/Loop-Core-Engineering/blob/main/ECOSYSTEM.md)
+
+---
+
+## Project layout
+
+| Path | Purpose |
+|------|---------|
+| [`loopgym/`](loopgym/) | Registry, envs, runtime, evaluators |
+| [`envs/loopbench/`](envs/loopbench/) | Task fixtures + LSS specs |
+| [`docs/api.md`](docs/api.md) | API reference |
+| [`examples/quickstart.py`](examples/quickstart.py) | Onboarding smoke test |
+
+---
+
+## Citation
+
+```bibtex
+@software{loopgym2026,
+  title={LoopGym: OpenAI Gym for LSS-Defined Agent Loops},
+  author={Malpani, Kanak},
+  year={2026},
+  url={https://github.com/KanakMalpani/LoopGym}
+}
+```
+
+---
+
+<p align="center">
+  <sub>MIT · v0.1 · <a href="CONTRIBUTING.md">Contributing</a> · <a href="SECURITY.md">Security</a> · <a href="PUBLISHING.md">PyPI</a></sub>
+</p>

loopgym-0.1.0/SECURITY.md

@@ -0,0 +1,15 @@
+# Security Policy
+
+## Supported versions
+
+| Version | Supported |
+|---------|-----------|
+| 0.1.x   | Yes       |
+
+## Reporting
+
+Report privately via [GitHub Security Advisories](https://github.com/KanakMalpani/LoopGym/security/advisories/new).
+
+## LiveEnv
+
+`LiveEnv` executes real model API calls. Never commit API keys; use environment variables. SimEnv is the default for CI and untrusted specs.

loopgym-0.1.0/STATUS.md

@@ -0,0 +1,29 @@
+# Status
+
+| Field | Value |
+|-------|-------|
+| **Phase** | v0.1 shipped |
+| **Symbol** | ✅ |
+| **Started** | 2026-06-13 |
+| **Shipped** | 2026-06-13 |
+| **Owner** | — |
+| **Blockers** | — |
+| **Notes** | Published at https://github.com/KanakMalpani/LoopGym |
+
+## Completion checklist
+
+- [x] `loopgym/` Python package
+- [x] `envs/loopbench/` stub envs (3 tasks)
+- [x] `examples/quickstart.py`
+- [x] `docs/api.md`
+- [x] ReplayEnv ↔ LoopNet seed integration
+- [x] CI: test workflow + LSS validation
+- [x] `SYNC.md` — canonical source policy
+- [ ] PyPI publish (`pip install loopgym` — pending trusted publisher or token on PyPI)
+
+## Links
+
+- Parent workspace: [../README.md](../README.md)
+- Core specs: [../01-loop-engineering-core/](../01-loop-engineering-core/)
+- Agent brief: [../AGENT-BRIEF.md](../AGENT-BRIEF.md)
+- Discipline repo (source runtime): https://github.com/KanakMalpani/Loop-Engineering

loopgym-0.1.0/SYNC.md

@@ -0,0 +1,35 @@
+# Sync policy — LoopGym
+
+**Runtime layer for Loop Engineering.**
+
+| Artifact | Canonical source | This repo |
+|----------|------------------|-----------|
+| LSS / LES specs | [Loop Core Engineering](https://github.com/KanakMalpani/Loop-Core-Engineering) | pin `lss@1.0.0`, `les@1.0.0` |
+| LoopNet records | [LoopNet](https://github.com/KanakMalpani/loopnet) | ReplayEnv reads seed JSONL |
+| Env IDs | Loop Core Engineering `specs/loop-ids.md` | registry in `loopgym/registry.py` |
+| Benchmark tasks | [LoopBench](https://github.com/KanakMalpani/LoopBench) | bundled fixtures under `envs/loopbench/` |
+
+**Repository:** https://github.com/KanakMalpani/LoopGym
+
+## Validation before release
+
+```bash
+pytest tests/ -q
+python examples/quickstart.py
+```
+
+Validate bundled specs against Loop Core Engineering in CI.
+
+## Do not duplicate
+
+- LSS JSON Schema — checkout Loop Core Engineering in CI
+- LoopBench task YAML — reference env IDs only
+- LES observed scoring — belongs in LoopBench
+
+## CI dependency order
+
+```
+Loop Core Engineering (schema)
+LoopGym (this repo)
+LoopBench (downstream)
+```