[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-1442":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":13,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":9,"pushedAt":9,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":14,"starSnapshotCount":14,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},1442,"dspy-agent-skills","intertwine\u002Fdspy-agent-skills","intertwine","Production-grade DSPy 3.2.x agent skills + validated end-to-end examples for Claude Code and Codex CLI — fundamentals, evaluation, GEPA, BetterTogether, and RLM.",null,"Python",249,22,1,0,3,4,23,9,54.39,"MIT License",false,"main",true,[],"2026-06-12 04:00:09","# DSPy Agent Skills\n\n[![DSPy 3.2.x](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDSPy-3.2.x-0A7B83)](https:\u002F\u002Fdspy.ai\u002F)\n\n**Production-grade DSPy 3.2.x skills for coding agents.** A synthesized, spec-compliant pack of five agent skills that turns Claude Code, Codex CLI, and any other [agentskills.io](https:\u002F\u002Fagentskills.io)-compatible agent into a DSPy expert.\n\n- ✅ Validated against DSPy 3.2.0 (the real API, not inferred from stale docs)\n- ✅ Single source of truth for both **Claude Code** and **Codex CLI**\n- ✅ Progressive disclosure (short `SKILL.md` + deep `reference.md`)\n- ✅ Runnable `example_*.py` scripts with offline `--dry-run`\n- ✅ Includes a DSPy 3.2.0 `BetterTogether` chaining example\n- ✅ Plugin manifest + marketplace manifest for one-click install\n- ✅ 80 validation tests (frontmatter spec, JSON schema, Python AST, skill-doc correctness guards)\n\n## What's inside\n\n| Skill | When it auto-invokes |\n|---|---|\n| [`dspy-fundamentals`](skills\u002Fdspy-fundamentals\u002FSKILL.md) | Any new DSPy code: Signatures, Modules, Predict\u002FChainOfThought\u002FReAct, save\u002Fload |\n| [`dspy-evaluation-harness`](skills\u002Fdspy-evaluation-harness\u002FSKILL.md) | Writing metrics, splitting dev\u002Fval sets, calling `dspy.Evaluate` |\n| [`dspy-gepa-optimizer`](skills\u002Fdspy-gepa-optimizer\u002FSKILL.md) | Optimizing\u002Fcompiling DSPy programs with `dspy.GEPA` |\n| [`dspy-rlm-module`](skills\u002Fdspy-rlm-module\u002FSKILL.md) | Long context, codebase QA, recursive exploration via `dspy.RLM` |\n| [`dspy-advanced-workflow`](skills\u002Fdspy-advanced-workflow\u002FSKILL.md) | End-to-end builds — orchestrates the other four |\n\n## Install\n\n### Claude Code (via marketplace)\n\n```text\n\u002Fplugin marketplace add intertwine\u002Fdspy-agent-skills\n\u002Fplugin install dspy-agent-skills@dspy-agent-skills\n```\n\n### Agent Skills CLI (`npx skills`)\n\n```bash\nnpx skills add intertwine\u002Fdspy-agent-skills --list\nnpx skills add intertwine\u002Fdspy-agent-skills --skill '*' -a codex -y\n```\n\nThe Vercel `skills` CLI currently expects a GitHub `owner\u002Frepo`, URL, well-known HTTPS endpoint, or local path as its source. The bare form `npx skills add dspy-agent-skills` is not resolvable unless the upstream CLI adds a source alias, so use `intertwine\u002Fdspy-agent-skills`.\n\n### Claude Code + Codex (repo checkout)\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fintertwine\u002Fdspy-agent-skills\ncd dspy-agent-skills\n.\u002Fscripts\u002Finstall.sh           # symlinks into ~\u002F.claude\u002Fskills\u002F and ~\u002F.agents\u002Fskills\u002F\n```\n\nFlags: `--claude-only`, `--codex-only`, `--copy` (copy instead of symlink), `--uninstall`, `--dry-run`.\n\n### Manual\n\nDrop `skills\u002F*` into `~\u002F.claude\u002Fskills\u002F` (Claude Code) or `~\u002F.agents\u002Fskills\u002F` (Codex CLI). See [docs\u002Finstallation.md](docs\u002Finstallation.md) for all options.\n\n## Five-second demo\n\nIn your agent, say:\n\n> \"Build a DSPy sentiment classifier, optimize it with GEPA, and save the artifact.\"\n\nThe agent auto-loads `dspy-advanced-workflow`, which chains the other skills and outputs a full baseline → GEPA → export pipeline. No further prompting needed.\n\n## End-to-end examples (current committed artifacts)\n\nThree runnable demos under [`examples\u002F`](examples\u002F) exercise every skill against real LMs and ship with **committed baseline vs. GEPA-optimized numbers** plus explicit `3.1.3` vs. `3.2.0` comparison notes.\n\n| Example | Artifact DSPy | Task LM | Baseline | Optimized | Δ | Status |\n|---|---|---|---:|---:|---:|---|\n| [01-rag-qa](examples\u002F01-rag-qa\u002F) | 3.2.0 | Ministral 3B 2512 | 80.47 | **100.00** | **+19.53** | Clean comparison refreshed on 2026-04-28 |\n| [02-math-reasoning](examples\u002F02-math-reasoning\u002F) | 3.2.0 | Ministral 3B 2512 | 85.00 | **93.33** | **+8.33** | Refreshed on 2026-04-21 |\n| [03-invoice-extraction](examples\u002F03-invoice-extraction\u002F) | 3.1.3 | Liquid LFM 2.5 1.2B (free) | 0.833 | **0.931** | **+0.098** | Historical artifact retained |\n\nThe refreshed `01` and `02` artifacts use the paid pair `openrouter\u002Fmistralai\u002Fministral-3b-2512` + `openrouter\u002Fqwen\u002Fqwen3-30b-a3b-instruct-2507`. `03` stays on its historical DSPy `3.1.3` artifact because a clean DSPy `3.2.0` baseline on the same Liquid\u002FNemotron pair already reached `0.944`, leaving little useful headroom for a replacement GEPA artifact. See [`examples\u002FREADME.md`](examples\u002FREADME.md) and each example's `version_comparison.md` for the exact commands and caveats.\n\n## Grounding\n\nEvery API claim is grounded in:\n\n- https:\u002F\u002Fdspy.ai\u002F (official docs, DSPy 3.2.x)\n- https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fskills.md (Claude Code skill spec)\n- https:\u002F\u002Fdevelopers.openai.com\u002Fcodex\u002Fskills (Codex skill spec)\n\n## Development\n\n```bash\n# Run validation suite\nuv run --with pytest python -m pytest tests\u002F -v\n\n# Smoke-test every example offline (no API key needed)\nfor f in skills\u002F*\u002Fexample_*.py; do uv run --with dspy python \"$f\" --dry-run; done\n\n# Live GEPA run (requires OPENAI_API_KEY)\ncd skills\u002Fdspy-advanced-workflow\nOPENAI_API_KEY=... uv run --with dspy python example_pipeline.py --auto light\n```\n\nIf `uv run --with dspy` resolves DSPy `3.1.3` instead of `3.2.0`, check whether `UV_EXCLUDE_NEWER` or a stale package mirror is hiding the new release. The exact 3.2.0 override we validated for this repo is:\n\n```bash\nenv -u UV_EXCLUDE_NEWER uv run --with dspy==3.2.0 python -c 'import dspy; print(dspy.__version__)'\n```\n\n## Compatibility\n\n- **DSPy**: 3.2.x (tested against 3.2.0)\n- **Claude Code**: current (skill spec as of 2026-04-17)\n- **Codex CLI**: current Agent Skills format\n- **Python**: 3.10+\n- **Deno**: required only for `dspy.RLM` examples (Pyodide sandbox)\n\n## Layout\n\n```\ndspy-agent-skills\u002F\n├── .claude-plugin\u002F\n│   ├── plugin.json\n│   └── marketplace.json\n├── skills\u002F\n│   ├── dspy-fundamentals\u002F{SKILL.md, reference.md, example_qa.py}\n│   ├── dspy-evaluation-harness\u002F{SKILL.md, reference.md, example_metric.py}\n│   ├── dspy-gepa-optimizer\u002F{SKILL.md, reference.md, example_gepa.py}\n│   ├── dspy-rlm-module\u002F{SKILL.md, reference.md, example_rlm.py}\n│   └── dspy-advanced-workflow\u002F{SKILL.md, example_pipeline.py}\n├── scripts\u002Finstall.sh           # dual-target installer\n├── tests\u002F                       # spec validators\n├── docs\u002F{installation,usage,CHANGELOG}.md\n├── README.md  LICENSE  .gitignore\n```\n\n## Version\n\n**v0.2.1** • Targets DSPy 3.2.x\n\n## License\n\nMIT — see [LICENSE](LICENSE).\n\n## Credits\n\nDraft contributors: Bryan Young ([@intertwine](https:\u002F\u002Fgithub.com\u002Fintertwine)) with Grok (xAI).\nValidation, spec-alignment, and dual-agent packaging: Claude Opus 4.7, April 2026.\n","该项目提供了一套生产级的DSPy 3.2.x编程代理技能，包括五个经过验证的技能包，适用于Claude Code和Codex CLI等兼容agentskills.io的代理。其核心功能涵盖从基础到高级的工作流程，如签名、模块定义、预测链优化及递归探索等，并通过了80项严格的验证测试以确保与DSPy 3.2.0 API的完全兼容性。此外，项目还提供了详尽的文档说明与即插即用的安装脚本，便于开发者快速上手。适合需要构建复杂DSPy应用或希望提升现有代码质量的研发团队使用，在自动化编程任务中发挥重要作用。",2,"2026-06-11 02:43:46","CREATED_QUERY"]