{"resourceId":"doe-genesis-mission-35-agent-skills-national-labs-20261029","versions":[{"version":"external-1a8840ef1399d7586d36b7e0d55ff732f6806ec9fbde797e5d632d6381f9de96","resource":{"id":"doe-genesis-mission-35-agent-skills-national-labs-20261029","title":"DOE releases 35 Genesis Mission agent skills packaging national-lab know-how for Claude Code, Codex, Gemini CLI and other coding agents","organization":"U.S. Department of Energy (ModCon BASE team, co-led by PNNL and Berkeley Lab), reported by R&D World","sector":"Federal research infrastructure and national laboratories","geography":"United States","publishedAt":"29 October 2026 (R&D World)","publicationDate":"2026-10-29","eventDate":null,"sourceName":"R&D World","sourceLabel":"Julia Rock-Torcivia, \"DOE packages national lab know-how as 35 skills for Claude Code, Codex and Gemini CLI,\" R&D World, 29 Oct 2026","sourceUrl":"https://www.rdworldonline.com/does-first-shared-genesis-mission-tool-works-with-claude-code-codex-and-gemini-cli/","evidenceClass":"independent-reporting","outcomeClass":"emerging","topics":["developers-agents","infrastructure","governance-procurement"],"finding":"R&D World reports that DOE released 35 skills for AI coding agents as the first technical resources from ModCon and the American Science Cloud under the Genesis Mission. The skills follow the Agent Skills format (a folder built around a SKILL.md file). Twenty-three include Python or shell scripts; the other 12, including job-scheduling guides for the Frontier, Aurora and Perlmutter supercomputers, are text only. They cover HPC schedulers, data tools such as data cards, AI safety analysis, American Science Cloud clients such as Globus Compute, literature search and multi-agent systems. Documentation says they work with Claude Code, Codex, Cursor, OpenCode, Gemini CLI and LangChain Deep Agents, though individual skills may depend on client features. Researchers at seven national labs wrote them. The contribution guide permits disclosed AI-assisted work and requires a human reviewer on every pull request; eight of 69 commits list Claude or GitHub Copilot as co-author.","sledRelevance":"University research computing centers already write user guides for schedulers and data services; DOE has shown a way to publish that know-how so researchers' coding agents use it correctly. Universities with allocations on DOE systems can use these skills directly.","evidence":"Read the full article, which draws on the repository and its attribution files. No usage, accuracy or adoption data is reported, and the repository was not inspected directly.","architectureImplications":"Agent skills become a distributable layer of site-specific documentation and scripts sitting between general coding agents and HPC or data services.","governanceImplications":"Human review on every pull request and disclosure of AI-assisted contributions are concrete controls for shared agent instructions.","securityPrivacyImplications":"Skills that include runnable scripts need review before use; individual skills may depend on external tools.","caveats":"Trade press report; claims about compatibility come from project documentation and were not tested. No evidence yet that the skills improve agent accuracy. Federal terminology (SI) and the Genesis Mission context are U.S.-specific.","streamIds":["research"],"roles":{"sales":"Lead with making existing research computing know-how usable by researchers' coding agents. R&D World reported on 29 October that DOE released 35 Genesis Mission skills, including scheduler guides for Frontier, Aurora and Perlmutter, that work across several coding agents. University research computing directors could publish similar skills for their own clusters. Offer a bounded skills-authoring engagement. Do not claim the skills improve accuracy or productivity; no usage or evaluation data is reported.","engineering":"Frame fit around a client-neutral skills layer for site-specific HPC and data workflows. Each skill is a folder with a SKILL.md file and, in 23 of 35 cases, runnable scripts. A university version needs current scheduler documentation, tested job templates and a review process. Validate by running representative researcher requests through two different coding agents with and without the skills and comparing job-script errors, since compatibility claims come from project documentation.","delivery":"Treat delivery as maintaining agent skills like documentation, with code review. Work covers writing skills for local schedulers and data services, testing them across agents, and versioning them with system changes. Ownership sits with the research computing user-support team. Governance checkpoints mirror DOE's human review on every pull request and AI-contribution disclosure. Acceptance criteria are proposals, such as fewer failed job submissions from agent-generated scripts, not reported results."},"roleDetails":{"sales":{"customerProblem":"Researchers use general coding agents that do not know local cluster rules, producing broken job scripts and support tickets.","stakeholders":["Research computing director","HPC user-support lead","CIO","Research data management lead"],"discoveryQuestions":["Which coding agents do researchers use on your clusters?","What share of support tickets involve job-script errors?","Do you hold allocations on DOE systems?"],"valueHypothesis":"Published local skills may reduce agent-generated job errors and support load, if tested.","potentialEngagement":"Author and test a starter set of skills for the institution's scheduler and data services.","unsupportedClaims":["Accuracy or productivity gains","Uniform compatibility across all agents","Adoption by researchers"]},"engineering":{"fit":"Fits institutions whose researchers use coding agents on shared clusters.","architectureIntegration":"SKILL.md-based folders with scripts, distributed through a repository and loaded by multiple coding agents.","prerequisites":["Current scheduler and storage documentation","Tested job templates","Repository with review workflow"],"constraints":["Skills may depend on client-specific features","Local policies differ from DOE systems","No evaluation data"],"security":"Review runnable scripts before publication and avoid embedding credentials or internal hostnames.","proofOfValue":["Compare job-script error rates with and without skills","Test each skill on at least two agents","Track support tickets after release"]},"delivery":{"work":"Write, test and version skills for local systems; publish with documentation updates.","dependencies":["Scheduler and storage teams","Repository hosting","Agent tools researchers use"],"ownership":"Research computing user-support team.","skillsAdoption":"Support staff learn skill authoring; researchers learn to install and update skills.","governanceCheckpoints":["Human review on every change","Disclosure of AI-assisted contributions","Re-test after system upgrades"],"acceptanceCriteria":["Fewer failed agent-generated submissions over a term (proposal)","Every skill tested on two agents","Skills updated within an agreed time after system changes"],"risks":["Stale skills after system changes","Scripts with unsafe defaults","Researchers trusting agents without checks"]}},"retrievedAt":"2026-10-29T22:01:29-05:00","enrichedAt":"2026-10-29T22:01:33-05:00","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Could lower the barrier for researchers new to supercomputer job submission, but is not evaluated.","procurementImplications":"Skills are client-neutral, reducing lock-in to one coding agent; frontier AI companies are Genesis partners.","operatingModelImplications":"Research computing staff may maintain agent skills alongside user documentation.","peerReview":"unknown","sourceVerification":{"openedUrl":"https://www.rdworldonline.com/does-first-shared-genesis-mission-tool-works-with-claude-code-codex-and-gemini-cli/","referenceExcerpt":"permits AI-assisted work with disclosure and requires a human reviewer on every pull request","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}