AI Research Skills (Orchestra)
Orchestra-Research/AI-Research-SKILLs98 skills for AI research and ML engineering, plus an autoresearch orchestrator that runs a two-loop experiment and synthesis cycle.
With caveats
Broad access Depth 4/5
paper-context-resolver skill
A narrow helper skill that looks up one reproduction-critical detail in a paper, such as a dataset split, when the repo README leaves a gap.
Our verdict Recommended with caveats
A sensible, tightly scoped prompt, but it is a 2 KB helper that only makes sense inside the RigorPilot reproduction flow.
What the paper-context-resolver skill does
paper-context-resolver is a small helper inside the RigorPilot Skills pack. It is meant to be called by the main reproduction skill, not on its own. Its job is deliberately limited: after the agent has read a repository's README and files and still has one concrete unanswered question, this skill goes to the paper to answer just that question. The listed question types are dataset version or split, preprocessing and postprocessing details, evaluation protocol, checkpoint or model variant mapping, and critical runtime assumptions.
The bundled reference file sets a source order: a paper link the README itself provides, then the official project or paper page, then the arXiv or OpenReview record, with Google Scholar used only to locate a primary source. It states what not to do: no full paper summary, no novelty discussion, no related-work tour and no guessing at best settings. If the README and the paper disagree, the agent must not silently replace the README guidance; it records the conflict, says which source claims what, and keeps the README-first policy in the final report. The folder contains SKILL.md (about 2 KB), one reference file and an openai.yaml agent definition, with no scripts.
Good for
- Settling one ambiguous detail while reproducing a paper's code
- Documenting README-versus-paper conflicts instead of hiding them
Not for
- Summarizing or explaining a paper
- Standalone use without a reproduction task
Depth 1/5
A 2 KB SKILL.md with a single 1-page reference file and an agent YAML, no scripts or tests of its own.
Dependencies and costs
- Web access to paper pages (arXiv, OpenReview, project pages)
- Best used with the parent RigorPilot ai-research-reproduction skill
Security review
Review before installingA small Markdown-only helper skill for the RigorPilot reproduction workflow that resolves a narrow paper detail from web pages when a README is unclear. It has no scripts, no allowed-tools pre-approvals and no credential references, and it is meant to be called by the main reproduction skill.
Worth knowing
- Makes network requestsThe helper looks up specific paper details (dataset split, preprocessing, evaluation protocol) on paper pages such as arXiv, OpenReview or project sites when a repository README leaves a gap; there is no script, so the agent uses its own web tools.SKILL.md:3
What we checked
The skill folder (3 files) on Oct 29, 2026. We had every file and read every script.
NVIDIA SkillSpector static scan: 0 pattern hits. SkillSpector reported 0 hits. The folder holds SKILL.md, one Markdown reference and an openai.yaml with display metadata.
A review, not a certification. Skills change after we check them — read the files before installing.
License
MIT — commercial use allowed (repository license).
Can it run unattended?
Not really. It answers a single question on demand and has no loop, state or stop conditions.
Alternatives
Academic Research Skills (ARS)
Imbad0202/academic-research-skillsFive very detailed Claude Code skills for the academic pipeline: deep research, paper writing, peer review, revision and systematic-review screening.
With caveats
Check first Depth 5/5
Deep Research Skill
199-biotechnologies/claude-deep-research-skillA single research skill with an 8-phase pipeline, parallel search, source scoring, claim and citation checks, and report output in Markdown, HTML and PDF.
With caveats
Check first Depth 4/5
FAQ
What is paper-context-resolver?
It is a helper skill from the RigorPilot Skills pack that resolves a specific reproduction detail, such as a dataset split or evaluation protocol, from the primary paper when a repository README is unclear. It does not summarize papers.
How do I install paper-context-resolver?
Run npx skills add lllllllama/RigorPilot-Skills --all to get the companion skills, since the main README documents the single-skill install only for ai-research-reproduction. It is intended to be invoked by the main reproduction skill.
What happens if the paper disagrees with the README?
The skill tells the agent not to override the README silently. It records the conflict, states which source says what, and keeps the README-first approach in the final report.
Sources: GitHub repository; install counts from skills.sh via Best Skills (CC BY 4.0). Review based on the published SKILL.md.