Skip to content
Back to search

RigorPilot-Skills

README-first research reproduction skills with bounded execution, auditable evidence, and byte-preserving README annotations.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/lllllllama--rigorpilot-skills/badge.svg)](/listings/lllllllama--rigorpilot-skills)

HTML

<a href="/listings/lllllllama--rigorpilot-skills"><img src="/api/listings/lllllllama--rigorpilot-skills/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

The installer needs Node.js/npm. Tested with `skills@1.5.26` and Node 22.20.0; that installer requires Node ≥22.20.0. If you see `EBADENGINE`, check the requested version.

Install the self-contained reproduction skill:

npx skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction

Install all skills for the companion research entrypoints:

npx skills add lllllllama/RigorPilot-Skills --all

These commands use the [current GitHub repository name](docs/install-repository-slug.md).

Open the target repository in a Skills-capable agent, then ask:

> Use ai-research-reproduction: run the smallest README-documented evaluation, preserve the source and write evidence to repro_outputs/, plus an annotated copy beside the original README. Ask before large downloads or long training.

The main skill works alone; choose **all skills** for companion and leaf entrypoints. Your existing agent loads the skill. The standalone model runner is optional. [Client compatibility](references/client-compatibility-policy.md)

Recommended first run:

| Step | What it does | What it does not do | |---|---|---| | Plan | Lists README-backed `cmd-XX` candidates, selects the smallest trusted target, and returns a selection fingerprint | No target execution, installs, downloads, source edits or evidence writes | | Run | Executes a reviewed `command-id` bound to that plan fingerprint and writes evidence under the target repo by default | Setup/download commands cannot be selected; a changed plan fails before target execution | | Verify | Rechecks the retained evidence, README round trip, runtime state and current source snapshot | Does not rerun the target command |

Start with the `RIGORPILOT_README.md` reported in `source_adjacent_readme.path`, then follow its command and log links. If a conflicting file blocks the extra copy, that file stays intact; inspect `repro_outputs/SUMMARY.md` for the outcome and next action.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 281 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # RigorPilot Skills
  2. Run research repositories from their README, with bounded execution and auditable evidence.
  3. RigorPilot adds section-level results without rewriting the original README.
  4. Trusted reproduction is the default; candidate exploration requires explicit authorization.
  5. [English](README.md) · [简体中文](README.zh-CN.md)
  6. [![Skillselion Top 100](https://skillselion.com/badge/skills/lllllllama/rigorpilot-skills/paper-context-resolver.svg?award=1)](https://skillselion.com/skills/lllllllama/rigorpilot-skills/paper-context-resolver)
  7. <p>
  8. <a href="https://github.com/lllllllama/RigorPilot-Skills/actions/workflows/validate.yml"><img alt="CI" src="https://github.com/lllllllama/RigorPilot-Skills/actions/workflows/validate.yml/badge.svg"></a>
  9. <a href="https://skillselion.com/skills/lllllllama/rigorpilot-skills/ai-research-reproduction"><img alt="Listed on Skillselion" src="https://skillselion.com/badge/skills/lllllllama/rigorpilot-skills/ai-research-reproduction.svg"></a>
  10. <a href="https://skills.sh/lllllllama/rigorpilot-skills"><img alt="skills.sh installs" src="https://skills.sh/b/lllllllama/rigorpilot-skills"></a>
  11. <a href="https://github.com/lllllllama/RigorPilot-Skills/stargazers"><img alt="GitHub stars" src="https://img.shields.io/github/stars/lllllllama/RigorPilot-Skills?style=flat-square"></a>
  12. <a href="LICENSE"><img alt="MIT License" src="https://img.shields.io/badge/license-MIT-yellow?style=flat-square"></a>
  13. <a href="https://agentskills.io"><img alt="Agent Skills standard" src="https://img.shields.io/badge/Agent%20Skills-open%20standard-1f6feb?style=flat-square"></a>
  14. <img alt="platforms" src="https://img.shields.io/badge/Windows%20%7C%20Linux-supported-6f42c1?style=flat-square">
  15. <img alt="local regression" src="https://img.shields.io/badge/local%20regression-80%2F80%20passed-8250df?style=flat-square">
  16. <a href="benchmark_outputs/external_suite_latest.json"><img alt="historical external protocols" src="https://img.shields.io/badge/historical%20protocols-4%2F4%20passed-238636?style=flat-square"></a>
  17. </p>
  18. <p align="center">
  19. <a href="#examples"><strong>Real examples</strong></a> ·
  20. <a href="#quick-start"><strong>Install & use</strong></a> ·
  21. <a href="#skills"><strong>Skill index</strong></a> ·
  22. <a href="#validation"><strong>Validation</strong></a> ·
  23. <a href="docs/ENGINEERING_ROADMAP.md"><strong>Engineering roadmap</strong></a>
  24. </p>
  25. <a id="examples"></a>
  26. <a id="evidence"></a>
  27. ## 📄 Real repositories, inspectable results
  28. Original commands, prose, badges, images, videos and HTML stay in the source file.
  29. RigorPilot splits that file into sections and inserts one evidence-linked card per section.
  30. Removing its insertion blocks restores the retained original README byte for byte.
  31. Each card below opens a full annotated README **beside the original README in a
  32. retained repository checkout**. Supporting repository files are kept so relative
  33. links and media retain their original context.
  34. 🟢 selected checks passed · 🔵 not executed · ⚪ read only · 🟡 partial · 🔴 blocked · 🟣 decision needed.
  35. Green does not automatically mean paper-result reproduction; blue is not an execution failure.
  36. <table>
  37. <tr>
  38. <td align="center" width="50%">
  39. <a href="benchmark_outputs/showcases/micrograd/repo/RIGORPILOT_README.md"><img src="assets/showcase/external-micrograd.png" width="100%" alt="micrograd: recorded pytest execution and section-level evidence"/></a><br/>
  40. <b>micrograd · correctness checks</b><br/>
  41. <sub>🟢 2 tests passed in 7.62 s<br/>8 headings = 8 annotations · original bytes preserved</sub><br/>
  42. <a href="benchmark_outputs/showcases/micrograd/repo/RIGORPILOT_README.md">Open full RigorPilot README →</a>
  43. </td>
  44. <td align="center" width="50%">
  45. <a href="benchmark_outputs/showcases/mingpt/repo/RIGORPILOT_README.md"><img src="assets/showcase/external-mingpt.png" width="100%" alt="minGPT: target selection only, with no model download or execution"/></a><br/>
  46. <b>minGPT · selection boundary</b><br/>
  47. <sub>🔵 Test selected, not executed · no model download<br/>11 headings = 11 annotations · original bytes preserved</sub><br/>
  48. <a href="benchmark_outputs/showcases/mingpt/repo/RIGORPILOT_README.md">Open full RigorPilot README →</a>
  49. </td>
  50. </tr>
  51. <tr>
  52. <td align="center" width="50%">
  53. <a href="benchmark_outputs/showcases/pytorch-mnist/repo/mnist/RIGORPILOT_README.md"><img src="assets/showcase/external-pytorch-mnist.png" width="100%" alt="PyTorch MNIST: partial bounded training and captured loss"/></a><br/>
  54. <b>PyTorch MNIST · bounded startup</b><br/>
  55. <sub>🟡 Partial training · observed loss 0.038893<br/>1 heading = 1 annotation · original bytes preserved</sub><br/>
  56. <a href="benchmark_outputs/showcases/pytorch-mnist/repo/mnist/RIGORPILOT_README.md">Open full RigorPilot README →</a>
  57. </td>
  58. <td align="center" width="50%">
  59. <a href="benchmark_outputs/showcases/nanogpt-shakespeare/repo/RIGORPILOT_README.md"><img src="assets/showcase/external-nanogpt.png" width="100%" alt="nanoGPT Shakespeare: partial CPU training and captured train and validation losses"/></a><br/>
  60. <b>nanoGPT Shakespeare · bounded training</b><br/>
  61. <sub>🟡 Partial · train loss 4.1676 · validation loss 4.1649<br/>11 headings = 11 annotations · original bytes preserved</sub><br/>
  62. <a href="benchmark_outputs/showcases/nanogpt-shakespeare/repo/RIGORPILOT_README.md">Open full RigorPilot README →</a>
  63. </td>
  64. </tr>
  65. </table>
  66. [All four cases and upstream links](benchmark_outputs/EXTERNAL_REPRODUCTIONS.md) ·
  67. [Recorded suite](benchmark_outputs/external_suite_latest.json) ·
  68. [Case definitions](benchmarks/external_cases.json) · [Methodology](benchmarks/README.md)
  69. These are historical, commit-pinned deterministic runs: **4/4 case protocols**
  70. passed in `251.0 s`, with a peak workspace of `98.67 MiB` and `0` model API calls.
  71. The zero-API count applies only to that suite. Selection-only and partial cases
  72. are not completed evaluations, converged training or reproduced paper scores.
  73. New: [installed-skill micrograd trial](docs/FIRST_USE_ACCEPTANCE.md), with before/after command reports, a retained failed attempt and independent checks—not a model-quality comparison.
  74. Current acceptance: explicit named-skill and fresh-client natural-language AUTO both pass end to end; earlier failures remain retained and A/B/model uplift remain unrun. [Compact status](docs/ACCEPTANCE_STATUS.md) · [real-client evidence](docs/REAL_CLIENT_ACCEPTANCE.md)
  75. <a id="quick-start"></a>
  76. ## 🚀 Install and use
  77. The installer needs Node.js/npm. Tested with `skills@1.5.26` and Node 22.20.0;
  78. that installer requires Node ≥22.20.0. If you see `EBADENGINE`, check the requested version.
  79. Install the self-contained reproduction skill:
  80. ```bash
  81. npx skills add lllllllama/RigorPilot-Skills --skill ai-research-reproduction
  82. ```
  83. Install all skills for the companion research entrypoints:
  84. ```bash
  85. npx skills add lllllllama/RigorPilot-Skills --all
  86. ```
  87. These commands use the [current GitHub repository name](docs/install-repository-slug.md).
  88. Open the target repository in a Skills-capable agent, then ask:
  89. > Use ai-research-reproduction: run the smallest README-documented evaluation, preserve the source and write evidence to repro_outputs/, plus an annotated copy beside the original README. Ask before large downloads or long training.
  90. The main skill works alone; choose **all skills** for companion and leaf entrypoints.
  91. Your existing agent loads the skill. The standalone model runner is optional.
  92. [Client compatibility](references/client-compatibility-policy.md)
  93. Recommended first run:
  94. | Step | What it does | What it does not do |
  95. |---|---|---|
  96. | Plan | Lists README-backed `cmd-XX` candidates, selects the smallest trusted target, and returns a selection fingerprint | No target execution, installs, downloads, source edits or evidence writes |
  97. | Run | Executes a reviewed `command-id` bound to that plan fingerprint and writes evidence under the target repo by default | Setup/download commands cannot be selected; a changed plan fails before target execution |
  98. | Verify | Rechecks the retained evidence, README round trip, runtime state and current source snapshot | Does not rerun the target command |
  99. Start with the `RIGORPILOT_README.md` reported in `source_adjacent_readme.path`,
  100. then follow its command and log links. If a conflicting file blocks the extra
  101. copy, that file stays intact; inspect `repro_outputs/SUMMARY.md` for the outcome
  102. and next action.
  103. ## What it does—and does not do
  104. README → documented target → reviewed setup → bounded execution → verification → evidence.
  105. - Preserves source meaning; records assumptions, deviations, failures and blockers.
  106. - Records process state, logs and attempt lineage; supports explicit cancellation,
  107. recovery and retry through the persistent runtime.
  108. - Separates trusted reproduction from explicitly authorized, candidate-only exploration.
  109. - Checks execution criteria independently of the model's completion claim.
  110. This is **local execution, not an OS sandbox**. Approved commands can access the
  111. host and network; use trusted repositories. Resource admission and between-action
  112. budget checks are not hard OS quotas or subscription-balance monitoring.
  113. The optional model loop currently supports Anthropic Messages and reviewed command
  114. IDs, not unrestricted source repair. **This standalone runner has no successful
  115. live-model acceptance recorded yet**: three provider attempts returned HTTP 502. Other model profiles
  116. are metadata, not proof of working transports or equivalent model performance.
  117. [Runner and recovery contract](skills/ai-research-reproduction/references/agent-runner.md) ·
  118. [Implementation evidence and limits](docs/P0_P1_DELIVERY.md)
  119. <a id="skills"></a>
  120. ## 🎯 Skill index
  121. | Task | Skill |
  122. |---|---|
  123. | Reproduce from README commands | [`ai-research-reproduction`](skills/ai-research-reproduction/SKILL.md) |
  124. | Read-only repository analysis | [`analyze-project`](skills/analyze-project/SKILL.md) |
  125. | Prepare environment, data and weights | [`env-and-assets-bootstrap`](skills/env-and-assets-bootstrap/SKILL.md) |
  126. | Run documented inference or evaluation | [`minimal-run-and-audit`](skills/minimal-run-and-audit/SKILL.md) |
  127. | Start or verify training conservatively | [`run-train`](skills/run-train/SKILL.md) |
  128. | Diagnose before proposing a patch | [`safe-debug`](skills/safe-debug/SKILL.md) |
  129. | Coordinate authorized candidate exploration | [`ai-research-explore`](skills/ai-research-explore/SKILL.md) |
  130. | Implement a candidate change on an isolated branch | [`explore-code`](skills/explore-code/SKILL.md) |
  131. | Execute a bounded candidate experiment | [`explore-run`](skills/explore-run/SKILL.md) |
  132. Two helpers support orchestration: `repo-intake-and-plan` and `paper-context-resolver`.
  133. Exploration requires a durable `current_research` anchor and a frozen comparison
  134. contract. Candidate results never become trusted baseline results by declaration.
  135. [Routing](references/routing-policy.md) · [Research loop](references/research-thinking-loop.md) ·
  136. [Campaign inputs](skills/ai-research-explore/references/research-campaign-spec.md)
  137. ## 📦 Evidence bundle
  138. | Artifact | What to inspect |
  139. |---|---|
  140. | `repro_outputs/ANNOTATED_README.md` | Original README with inserted section verdicts |
  141. | `SUMMARY.md`, `COMMANDS.md`, `LOG.md`, `status.json` | Outcome, exact commands, reviewed selection provenance, stable `error.code`, observations and machine-readable status |
  142. | `invocation.json`, `evidence_manifest.json` | Invocation/source-integrity summary plus retained file sizes and SHA-256 hashes for local consistency checks |
  143. | `PATCHES.md`, `SCIENTIFIC_CHANGELOG.md`, `COMPARABILITY_REPORT.md` | Changes, scientific meaning and comparison boundaries |
  144. | `_runtime/<run_id>/` | Process state, events, resource samples and stdout/stderr |
  145. | `.repro_job/` | Optional short-call supervisor: frozen request, reusable receipt, phase logs and completion-time verification |
  146. | `agent_state.json`, `trajectory.jsonl` | Optional model runner's checkpoints, tool calls and reported usage |
  147. 🟢 success · 🔵 not executed · ⚪ read only · 🟡 partial · 🔴 blocked · 🟣 decision required
  148. Standard evidence stays under `repro_outputs/`. Both main runners accept
  149. `--source-adjacent-readme` to also write `RIGORPILOT_README.md` beside the original,
  150. preserving the context of its relative media/file links. Only inserted evidence
  151. links are rebased. The same output directory may refresh its unchanged owned
  152. copy, never an unrelated or manually edited file. Retain supporting repository
  153. files and the evidence directory's `readme_delivery.json`.
  154. [Output contract](references/output-contract.md) · [Rigor principles](references/research-rigor-principles.md)
  155. For agent clients, the shortest normal path is `--plan-only --agent-output`, then
  156. run a returned `cmd-XX` with its `--plan-fingerprint`, then `--verify-output --agent-output`.
  157. The fingerprint binds execution to the reviewed README command set; setup/download
  158. commands are excluded from reproduction candidates. Routine success should not
  159. require reading orchestrator, bundle, writer, or runtime source.
  160. When `--output-dir` is omitted, the orchestrator writes to the target repository's
  161. `repro_outputs/` directory, not the caller's current working directory.
  162. The evidence manifest detects later changes against the retained manifest; it is
  163. not a digital signature, external attestation or substitute for an OS sandbox.
  164. Short tool-call hosts can opt in with `--plan-only --include-agent-handoff` to
  165. receive identity-bound start/status/cancel argv. Ordinary plans omit this extra
  166. control path. [Short-call contract](skills/ai-research-reproduction/references/agent-job.md)
  167. <a id="validation"></a>
  168. ## ✅ Offline validation
  169. For normal development, run the high-signal core suite first:
  170. ```bash
  171. python -m pip install -r requirements-dev.txt
  172. python scripts/run_all_tests.py --core
  173. ```
  174. Before commit/release, run the complete automatically discovered suite:
  175. ```bash
  176. python scripts/run_all_tests.py
  177. python scripts/check_publication.py
  178. ```
  179. The latest version-bound counts and timing are in the
  180. [acceptance ledger](docs/ACCEPTANCE_STATUS.md). Reports and complete per-script
  181. logs are written outside the repository by default; use `--report PATH` to
  182. choose a location. The core suite does not replace the full suite or CI.
  183. For installed-layout functional acceptance without model calls:
  184. ```bash
  185. python benchmarks/run_skill_acceptance.py --output tmp/skill-check
  186. ```
  187. It covers positive/negative evidence behavior and can optionally include pinned
  188. micrograd when PyTorch/pytest already exist. [Functional acceptance](docs/SKILL_ACCEPTANCE.md)
  189. For installation/environment problems (replace the skill path after installation):
  190. ```bash
  191. python skills/ai-research-reproduction/scripts/doctor.py --repo /path/to/target
  192. ```
  193. This read-only check reports the actual Python, Git, bundled-file integrity and
  194. README availability. Add `--require-module torch --require-module pytest` for
  195. dependency discovery; it does not install anything or execute target code.
  196. Deeper checks are intentionally off the main path: [benchmark methodology](benchmarks/README.md) ·
  197. [agent runtime/fault evidence](docs/AGENT_RUNTIME_REVIEW.md) ·
  198. [paired A/B kit](docs/PAIRED_PILOT.md) · [controlled trials](docs/CONTROLLED_TRIALS.md).
  199. Local tests do not substitute for live-model or held-out evaluation.
  200. ## Engineering and contributions
  201. [Engineering roadmap](docs/ENGINEERING_ROADMAP.md) · [Contributing](CONTRIBUTING.md) ·
  202. [Security and reporting](SECURITY.md) · [CI workflow](.github/workflows/validate.yml) ·
  203. [Reproduction feedback](https://github.com/lllllllama/RigorPilot-Skills/issues/new?template=reproduction.yml) ·
  204. [MIT license](LICENSE)
  205. Keep acceptance checks independent, retain failed evidence and review traces
  206. before publication. Do not publish credentials or unreviewed private repository data.
  207. [Agent guidance](AGENTS.md) · [Operating principles](references/agent-operating-principles.md) ·
  208. [Personalization policy](references/continuous-learning-policy.md)
  209. <details>
  210. <summary>Historical interface illustration—not execution evidence</summary>
  211. <img src="assets/annotated-readme-preview.png" width="840" alt="Historical MiniSeg interface illustration, not independently verified execution evidence"/>
  212. [First attempt](examples/annotated-readme-demo/first-run/ANNOTATED_README.md) ·
  213. [After setup](examples/annotated-readme-demo/after-setup/ANNOTATED_README.md).
  214. This older MiniSeg preview illustrates error, metric and authorization displays.
  215. Its execution provenance is not independently verified; it is excluded from benchmarks.
  216. </details>

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.