[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-preference-update":3,"mdc--i6lym6-key":33,"related-org-nvidia-preference-update":511,"related-repo-nvidia-preference-update":672},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":22,"repoUrl":23,"updatedAt":24,"license":25,"forks":26,"topics":27,"repo":28,"sourceUrl":31,"mdContent":32},"preference-update","update preference policies from user corrections","Update a bounded preference policy from the user's repeated corrections, without ever overwriting the user's own choices.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19],{"name":13,"slug":14,"type":15},"Agents","agents","tag",{"name":17,"slug":18,"type":15},"Personalization","personalization",{"name":20,"slug":21,"type":15},"Policy","policy",72,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnemoclaw-community","2026-08-31T09:19:26.384257","Apache-2.0",32,[],{"repoUrl":23,"stars":22,"forks":26,"topics":29,"description":30},[],"Our repository of community driven examples, showcases, and integrations","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnemoclaw-community\u002Ftree\u002FHEAD\u002Fexamples\u002Frecipes\u002Fnvidia\u002Fmemory-driven-chief-of-staff\u002Fprofile\u002Fskills\u002Fpreference-update","---\nname: preference-update\ndescription: Update a bounded preference policy from the user's repeated corrections, without ever overwriting the user's own choices.\nversion: 0.1.0\nlicense: Apache-2.0\nplatforms: [linux]\nmetadata:\n  hermes:\n    tags: [memory, policy, preferences]\n---\n\n# Updating preferences from corrections\n\nEvery time the user ignores a row or overrides its priority, they are\ncorrecting a judgment. One correction is an accident. Three of the same shape\nis a preference, and a preference belongs in writing where the next run can\nread it.\n\nNothing here trains a model. The output is a bounded text policy that later\nruns read as input — it can be inspected, edited by hand, and deleted, and\ndeleting it returns the system to its default judgment.\n\nThis skill reads the audit trail and edits one file. It never touches an\nobligation.\n\n## The invariant\n\n**Never write to `obligations`.** Not `priority`, not `manual_priority`, not\n`status`. The user's overrides are theirs; this skill only reads them.\nA run that changes a row has failed, however good its reasoning was.\n\nThe policy lands in `$HERMES_HOME\u002Fworkspace\u002Fpolicy\u002Fpreferences.md`, which\n`inbound-judging` reads as a prior and `obligation-review` reads as context.\n\n## Step 1 — collect\n\nRead the audit trail for user-authored corrections since the last run:\n\n```sql\nSELECT e.event_type, e.after_json, o.title, o.kind, i.source, i.sender\n  FROM events e\n  JOIN obligations o ON o.id = e.obligation_id\n  JOIN items i ON i.source_id = o.source_id\n WHERE e.actor = 'user'\n   AND e.event_type IN ('ignored', 'priority_override')\n   AND e.ts > :last_run\n```\n\nNote that `actor` matters. Rows this agent changed are not corrections;\nreading your own output back as evidence is how a system talks itself into a\nbelief.\n\n## Step 2 — group\n\nGroup the corrections by what they have in common. Useful groupings, roughly\nin order of how often they turn out to be real:\n\n- the same sender, or the same sender domain,\n- the same recurring subject shape — a build notification, a newsletter, an\n  automated digest,\n- the same `kind` from the same source.\n\nA pattern is one plain sentence that would describe most rows in its group. If\nyou cannot write that sentence without listing exceptions, it is not a pattern.\n\n## Step 3 — apply the threshold\n\n**A pattern qualifies only when it covers at least three corrections.** Below\nthat, leave it alone and let the next run decide — two of anything is a\ncoincidence you would be encoding forever.\n\nThe threshold is deliberately fixed. A system permitted to lower its own bar\nfor what counts as a preference eventually accepts everything.\n\n## Step 4 — write\n\nRead `$HERMES_HOME\u002Fworkspace\u002Fpolicy\u002Fpreferences.md`, creating it with an empty\n`## Observed Preferences` section if absent.\n\nIf an existing entry already says the same thing — the same pattern in other\nwords counts — do nothing. Restating a known pattern inflates the file and\nteaches the reader nothing.\n\nOtherwise append under `## Observed Preferences`, newest first:\n\n```markdown\n- [2026-08-18] Automated build notifications from ci@example.com — the user\n  ignores these; do not create a row unless a human is named in the body.\n```\n\nEach entry states the pattern and what to do about it. \"Deprioritize\nnewsletters\" is not actionable; \"newsletters from the vendor list are\nreview-only unless they name a deadline\" is.\n\n**Cap the section at 20 entries.** Past that, drop the oldest. A policy file\nlonger than one screen stops being read, and an unread policy is worse than\nnone — it implies the system is adapting when it is not.\n\nRead the file back after editing and confirm both that the entry landed and\nthat the cap held.\n\n## Step 5 — log\n\nAppend one entry to `$HERMES_HOME\u002Fworkspace\u002Fmemory\u002Flog.md`, including runs that changed\nnothing:\n\n```markdown\n## [2026-08-18T09:20:00Z] preference-update\n- 7 user corrections since last run, 2 patterns above threshold, 1 already known\n- appended: automated build notifications from ci@example.com\n```\n\nA run that changed nothing is a useful signal — it means the judging skill is\ncurrently matching the user's taste.\n",{"data":34,"body":43},{"name":4,"description":6,"version":35,"license":25,"platforms":36,"metadata":38},"0.1.0",[37],"linux",{"hermes":39},{"tags":40},[41,21,42],"memory","preferences",{"type":44,"children":45},"root",[46,55,61,66,71,78,122,151,157,162,238,251,257,262,290,295,301,311,316,322,342,347,359,407,412,422,427,433,446,500,505],{"type":47,"tag":48,"props":49,"children":51},"element","h1",{"id":50},"updating-preferences-from-corrections",[52],{"type":53,"value":54},"text","Updating preferences from corrections",{"type":47,"tag":56,"props":57,"children":58},"p",{},[59],{"type":53,"value":60},"Every time the user ignores a row or overrides its priority, they are\ncorrecting a judgment. One correction is an accident. Three of the same shape\nis a preference, and a preference belongs in writing where the next run can\nread it.",{"type":47,"tag":56,"props":62,"children":63},{},[64],{"type":53,"value":65},"Nothing here trains a model. The output is a bounded text policy that later\nruns read as input — it can be inspected, edited by hand, and deleted, and\ndeleting it returns the system to its default judgment.",{"type":47,"tag":56,"props":67,"children":68},{},[69],{"type":53,"value":70},"This skill reads the audit trail and edits one file. It never touches an\nobligation.",{"type":47,"tag":72,"props":73,"children":75},"h2",{"id":74},"the-invariant",[76],{"type":53,"value":77},"The invariant",{"type":47,"tag":56,"props":79,"children":80},{},[81,96,98,104,106,112,114,120],{"type":47,"tag":82,"props":83,"children":84},"strong",{},[85,87,94],{"type":53,"value":86},"Never write to ",{"type":47,"tag":88,"props":89,"children":91},"code",{"className":90},[],[92],{"type":53,"value":93},"obligations",{"type":53,"value":95},".",{"type":53,"value":97}," Not ",{"type":47,"tag":88,"props":99,"children":101},{"className":100},[],[102],{"type":53,"value":103},"priority",{"type":53,"value":105},", not ",{"type":47,"tag":88,"props":107,"children":109},{"className":108},[],[110],{"type":53,"value":111},"manual_priority",{"type":53,"value":113},", not\n",{"type":47,"tag":88,"props":115,"children":117},{"className":116},[],[118],{"type":53,"value":119},"status",{"type":53,"value":121},". The user's overrides are theirs; this skill only reads them.\nA run that changes a row has failed, however good its reasoning was.",{"type":47,"tag":56,"props":123,"children":124},{},[125,127,133,135,141,143,149],{"type":53,"value":126},"The policy lands in ",{"type":47,"tag":88,"props":128,"children":130},{"className":129},[],[131],{"type":53,"value":132},"$HERMES_HOME\u002Fworkspace\u002Fpolicy\u002Fpreferences.md",{"type":53,"value":134},", which\n",{"type":47,"tag":88,"props":136,"children":138},{"className":137},[],[139],{"type":53,"value":140},"inbound-judging",{"type":53,"value":142}," reads as a prior and ",{"type":47,"tag":88,"props":144,"children":146},{"className":145},[],[147],{"type":53,"value":148},"obligation-review",{"type":53,"value":150}," reads as context.",{"type":47,"tag":72,"props":152,"children":154},{"id":153},"step-1-collect",[155],{"type":53,"value":156},"Step 1 — collect",{"type":47,"tag":56,"props":158,"children":159},{},[160],{"type":53,"value":161},"Read the audit trail for user-authored corrections since the last run:",{"type":47,"tag":163,"props":164,"children":169},"pre",{"className":165,"code":166,"language":167,"meta":168,"style":168},"language-sql shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","SELECT e.event_type, e.after_json, o.title, o.kind, i.source, i.sender\n  FROM events e\n  JOIN obligations o ON o.id = e.obligation_id\n  JOIN items i ON i.source_id = o.source_id\n WHERE e.actor = 'user'\n   AND e.event_type IN ('ignored', 'priority_override')\n   AND e.ts > :last_run\n","sql","",[170],{"type":47,"tag":88,"props":171,"children":172},{"__ignoreMap":168},[173,184,193,202,211,220,229],{"type":47,"tag":174,"props":175,"children":178},"span",{"class":176,"line":177},"line",1,[179],{"type":47,"tag":174,"props":180,"children":181},{},[182],{"type":53,"value":183},"SELECT e.event_type, e.after_json, o.title, o.kind, i.source, i.sender\n",{"type":47,"tag":174,"props":185,"children":187},{"class":176,"line":186},2,[188],{"type":47,"tag":174,"props":189,"children":190},{},[191],{"type":53,"value":192},"  FROM events e\n",{"type":47,"tag":174,"props":194,"children":196},{"class":176,"line":195},3,[197],{"type":47,"tag":174,"props":198,"children":199},{},[200],{"type":53,"value":201},"  JOIN obligations o ON o.id = e.obligation_id\n",{"type":47,"tag":174,"props":203,"children":205},{"class":176,"line":204},4,[206],{"type":47,"tag":174,"props":207,"children":208},{},[209],{"type":53,"value":210},"  JOIN items i ON i.source_id = o.source_id\n",{"type":47,"tag":174,"props":212,"children":214},{"class":176,"line":213},5,[215],{"type":47,"tag":174,"props":216,"children":217},{},[218],{"type":53,"value":219}," WHERE e.actor = 'user'\n",{"type":47,"tag":174,"props":221,"children":223},{"class":176,"line":222},6,[224],{"type":47,"tag":174,"props":225,"children":226},{},[227],{"type":53,"value":228},"   AND e.event_type IN ('ignored', 'priority_override')\n",{"type":47,"tag":174,"props":230,"children":232},{"class":176,"line":231},7,[233],{"type":47,"tag":174,"props":234,"children":235},{},[236],{"type":53,"value":237},"   AND e.ts > :last_run\n",{"type":47,"tag":56,"props":239,"children":240},{},[241,243,249],{"type":53,"value":242},"Note that ",{"type":47,"tag":88,"props":244,"children":246},{"className":245},[],[247],{"type":53,"value":248},"actor",{"type":53,"value":250}," matters. Rows this agent changed are not corrections;\nreading your own output back as evidence is how a system talks itself into a\nbelief.",{"type":47,"tag":72,"props":252,"children":254},{"id":253},"step-2-group",[255],{"type":53,"value":256},"Step 2 — group",{"type":47,"tag":56,"props":258,"children":259},{},[260],{"type":53,"value":261},"Group the corrections by what they have in common. Useful groupings, roughly\nin order of how often they turn out to be real:",{"type":47,"tag":263,"props":264,"children":265},"ul",{},[266,272,277],{"type":47,"tag":267,"props":268,"children":269},"li",{},[270],{"type":53,"value":271},"the same sender, or the same sender domain,",{"type":47,"tag":267,"props":273,"children":274},{},[275],{"type":53,"value":276},"the same recurring subject shape — a build notification, a newsletter, an\nautomated digest,",{"type":47,"tag":267,"props":278,"children":279},{},[280,282,288],{"type":53,"value":281},"the same ",{"type":47,"tag":88,"props":283,"children":285},{"className":284},[],[286],{"type":53,"value":287},"kind",{"type":53,"value":289}," from the same source.",{"type":47,"tag":56,"props":291,"children":292},{},[293],{"type":53,"value":294},"A pattern is one plain sentence that would describe most rows in its group. If\nyou cannot write that sentence without listing exceptions, it is not a pattern.",{"type":47,"tag":72,"props":296,"children":298},{"id":297},"step-3-apply-the-threshold",[299],{"type":53,"value":300},"Step 3 — apply the threshold",{"type":47,"tag":56,"props":302,"children":303},{},[304,309],{"type":47,"tag":82,"props":305,"children":306},{},[307],{"type":53,"value":308},"A pattern qualifies only when it covers at least three corrections.",{"type":53,"value":310}," Below\nthat, leave it alone and let the next run decide — two of anything is a\ncoincidence you would be encoding forever.",{"type":47,"tag":56,"props":312,"children":313},{},[314],{"type":53,"value":315},"The threshold is deliberately fixed. A system permitted to lower its own bar\nfor what counts as a preference eventually accepts everything.",{"type":47,"tag":72,"props":317,"children":319},{"id":318},"step-4-write",[320],{"type":53,"value":321},"Step 4 — write",{"type":47,"tag":56,"props":323,"children":324},{},[325,327,332,334,340],{"type":53,"value":326},"Read ",{"type":47,"tag":88,"props":328,"children":330},{"className":329},[],[331],{"type":53,"value":132},{"type":53,"value":333},", creating it with an empty\n",{"type":47,"tag":88,"props":335,"children":337},{"className":336},[],[338],{"type":53,"value":339},"## Observed Preferences",{"type":53,"value":341}," section if absent.",{"type":47,"tag":56,"props":343,"children":344},{},[345],{"type":53,"value":346},"If an existing entry already says the same thing — the same pattern in other\nwords counts — do nothing. Restating a known pattern inflates the file and\nteaches the reader nothing.",{"type":47,"tag":56,"props":348,"children":349},{},[350,352,357],{"type":53,"value":351},"Otherwise append under ",{"type":47,"tag":88,"props":353,"children":355},{"className":354},[],[356],{"type":53,"value":339},{"type":53,"value":358},", newest first:",{"type":47,"tag":163,"props":360,"children":364},{"className":361,"code":362,"language":363,"meta":168,"style":168},"language-markdown shiki shiki-themes material-theme-lighter material-theme material-theme-palenight","- [2026-08-18] Automated build notifications from ci@example.com — the user\n  ignores these; do not create a row unless a human is named in the body.\n","markdown",[365],{"type":47,"tag":88,"props":366,"children":367},{"__ignoreMap":168},[368,399],{"type":47,"tag":174,"props":369,"children":370},{"class":176,"line":177},[371,377,382,388,393],{"type":47,"tag":174,"props":372,"children":374},{"style":373},"--shiki-light:#39ADB5;--shiki-default:#89DDFF;--shiki-dark:#89DDFF",[375],{"type":53,"value":376},"-",{"type":47,"tag":174,"props":378,"children":379},{"style":373},[380],{"type":53,"value":381}," [",{"type":47,"tag":174,"props":383,"children":385},{"style":384},"--shiki-light:#91B859;--shiki-default:#C3E88D;--shiki-dark:#C3E88D",[386],{"type":53,"value":387},"2026-08-18",{"type":47,"tag":174,"props":389,"children":390},{"style":373},[391],{"type":53,"value":392},"]",{"type":47,"tag":174,"props":394,"children":396},{"style":395},"--shiki-light:#90A4AE;--shiki-default:#EEFFFF;--shiki-dark:#BABED8",[397],{"type":53,"value":398}," Automated build notifications from ci@example.com — the user\n",{"type":47,"tag":174,"props":400,"children":401},{"class":176,"line":186},[402],{"type":47,"tag":174,"props":403,"children":404},{"style":395},[405],{"type":53,"value":406},"  ignores these; do not create a row unless a human is named in the body.\n",{"type":47,"tag":56,"props":408,"children":409},{},[410],{"type":53,"value":411},"Each entry states the pattern and what to do about it. \"Deprioritize\nnewsletters\" is not actionable; \"newsletters from the vendor list are\nreview-only unless they name a deadline\" is.",{"type":47,"tag":56,"props":413,"children":414},{},[415,420],{"type":47,"tag":82,"props":416,"children":417},{},[418],{"type":53,"value":419},"Cap the section at 20 entries.",{"type":53,"value":421}," Past that, drop the oldest. A policy file\nlonger than one screen stops being read, and an unread policy is worse than\nnone — it implies the system is adapting when it is not.",{"type":47,"tag":56,"props":423,"children":424},{},[425],{"type":53,"value":426},"Read the file back after editing and confirm both that the entry landed and\nthat the cap held.",{"type":47,"tag":72,"props":428,"children":430},{"id":429},"step-5-log",[431],{"type":53,"value":432},"Step 5 — log",{"type":47,"tag":56,"props":434,"children":435},{},[436,438,444],{"type":53,"value":437},"Append one entry to ",{"type":47,"tag":88,"props":439,"children":441},{"className":440},[],[442],{"type":53,"value":443},"$HERMES_HOME\u002Fworkspace\u002Fmemory\u002Flog.md",{"type":53,"value":445},", including runs that changed\nnothing:",{"type":47,"tag":163,"props":447,"children":449},{"className":361,"code":448,"language":363,"meta":168,"style":168},"## [2026-08-18T09:20:00Z] preference-update\n- 7 user corrections since last run, 2 patterns above threshold, 1 already known\n- appended: automated build notifications from ci@example.com\n",[450],{"type":47,"tag":88,"props":451,"children":452},{"__ignoreMap":168},[453,476,488],{"type":47,"tag":174,"props":454,"children":455},{"class":176,"line":177},[456,461,466,470],{"type":47,"tag":174,"props":457,"children":458},{"style":373},[459],{"type":53,"value":460},"## [",{"type":47,"tag":174,"props":462,"children":463},{"style":384},[464],{"type":53,"value":465},"2026-08-18T09:20:00Z",{"type":47,"tag":174,"props":467,"children":468},{"style":373},[469],{"type":53,"value":392},{"type":47,"tag":174,"props":471,"children":473},{"style":472},"--shiki-light:#E2931D;--shiki-default:#FFCB6B;--shiki-dark:#FFCB6B",[474],{"type":53,"value":475}," preference-update\n",{"type":47,"tag":174,"props":477,"children":478},{"class":176,"line":186},[479,483],{"type":47,"tag":174,"props":480,"children":481},{"style":373},[482],{"type":53,"value":376},{"type":47,"tag":174,"props":484,"children":485},{"style":395},[486],{"type":53,"value":487}," 7 user corrections since last run, 2 patterns above threshold, 1 already known\n",{"type":47,"tag":174,"props":489,"children":490},{"class":176,"line":195},[491,495],{"type":47,"tag":174,"props":492,"children":493},{"style":373},[494],{"type":53,"value":376},{"type":47,"tag":174,"props":496,"children":497},{"style":395},[498],{"type":53,"value":499}," appended: automated build notifications from ci@example.com\n",{"type":47,"tag":56,"props":501,"children":502},{},[503],{"type":53,"value":504},"A run that changed nothing is a useful signal — it means the judging skill is\ncurrently matching the user's taste.",{"type":47,"tag":506,"props":507,"children":508},"style",{},[509],{"type":53,"value":510},"html .light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html.light .shiki span {color: var(--shiki-light);background: var(--shiki-light-bg);font-style: var(--shiki-light-font-style);font-weight: var(--shiki-light-font-weight);text-decoration: var(--shiki-light-text-decoration);}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}",{"items":512,"total":671},[513,531,549,560,572,586,599,613,626,637,651,660],{"slug":514,"name":514,"fn":515,"description":516,"org":517,"tags":518,"stars":528,"repoUrl":529,"updatedAt":530},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[519,522,525],{"name":520,"slug":521,"type":15},"Documentation","documentation",{"name":523,"slug":524,"type":15},"MCP","mcp",{"name":526,"slug":527,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-08-25T03:29:57.273192",{"slug":532,"name":532,"fn":533,"description":534,"org":535,"tags":536,"stars":546,"repoUrl":547,"updatedAt":548},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[537,540,543],{"name":538,"slug":539,"type":15},"Containers","containers",{"name":541,"slug":542,"type":15},"Deployment","deployment",{"name":544,"slug":545,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":550,"name":550,"fn":551,"description":552,"org":553,"tags":554,"stars":546,"repoUrl":547,"updatedAt":559},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[555,558],{"name":556,"slug":557,"type":15},"CI\u002FCD","ci-cd",{"name":541,"slug":542,"type":15},"2026-07-14T05:25:59.97109",{"slug":561,"name":561,"fn":562,"description":563,"org":564,"tags":565,"stars":546,"repoUrl":547,"updatedAt":571},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[566,567,568],{"name":556,"slug":557,"type":15},{"name":541,"slug":542,"type":15},{"name":569,"slug":570,"type":15},"GitHub","github","2026-08-28T14:38:16.959248",{"slug":573,"name":573,"fn":574,"description":575,"org":576,"tags":577,"stars":546,"repoUrl":547,"updatedAt":585},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[578,581,582],{"name":579,"slug":580,"type":15},"Debugging","debugging",{"name":569,"slug":570,"type":15},{"name":583,"slug":584,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":587,"name":587,"fn":588,"description":589,"org":590,"tags":591,"stars":546,"repoUrl":547,"updatedAt":598},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[592,595],{"name":593,"slug":594,"type":15},"Best Practices","best-practices",{"name":596,"slug":597,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":600,"name":600,"fn":601,"description":602,"org":603,"tags":604,"stars":546,"repoUrl":547,"updatedAt":612},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel, including the mechanical steps for transferring an existing pretrain_gpt.py launch script.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[605,608,611],{"name":606,"slug":607,"type":15},"Machine Learning","machine-learning",{"name":609,"slug":610,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-08-21T03:36:57.470256",{"slug":614,"name":614,"fn":615,"description":616,"org":617,"tags":618,"stars":546,"repoUrl":547,"updatedAt":625},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[619,622],{"name":620,"slug":621,"type":15},"QA","qa",{"name":623,"slug":624,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":627,"name":627,"fn":628,"description":629,"org":630,"tags":631,"stars":546,"repoUrl":547,"updatedAt":636},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[632,633],{"name":541,"slug":542,"type":15},{"name":634,"slug":635,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":638,"name":638,"fn":639,"description":640,"org":641,"tags":642,"stars":546,"repoUrl":547,"updatedAt":650},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[643,646,647],{"name":644,"slug":645,"type":15},"Code Review","code-review",{"name":569,"slug":570,"type":15},{"name":648,"slug":649,"type":15},"Pull Requests","pull-requests","2026-08-25T03:29:16.211287",{"slug":652,"name":652,"fn":653,"description":654,"org":655,"tags":656,"stars":546,"repoUrl":547,"updatedAt":659},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[657,658],{"name":620,"slug":621,"type":15},{"name":623,"slug":624,"type":15},"2026-07-14T05:25:54.928983",{"slug":661,"name":661,"fn":662,"description":663,"org":664,"tags":665,"stars":546,"repoUrl":547,"updatedAt":670},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[666,669],{"name":667,"slug":668,"type":15},"Automation","automation",{"name":556,"slug":557,"type":15},"2026-07-30T05:29:03.275638",563,{"items":673,"total":774},[674,686,701,718,732,744,760],{"slug":675,"name":675,"fn":676,"description":677,"org":678,"tags":679,"stars":22,"repoUrl":23,"updatedAt":685},"axe-a11y","audit web accessibility with axe-core","Automated web accessibility scanning (axe-core) driven by patchright + real Google Chrome — audits sites behind bot walls and behind auth",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[680,683,684],{"name":681,"slug":682,"type":15},"Accessibility","accessibility",{"name":620,"slug":621,"type":15},{"name":623,"slug":624,"type":15},"2026-08-31T09:19:20.559156",{"slug":687,"name":687,"fn":688,"description":689,"org":690,"tags":691,"stars":22,"repoUrl":23,"updatedAt":700},"blackwall-payment-gate","screen and submit x402 payments","Screen x402 payments with an advisory Blackwall verdict, then submit a payment intent to the host-side release gate. You prepare payments; only the gate can settle them.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[692,695,698],{"name":693,"slug":694,"type":15},"Payments","payments",{"name":696,"slug":697,"type":15},"Security","security",{"name":699,"slug":699,"type":15},"x402","2026-08-31T09:19:25.64269",{"slug":702,"name":702,"fn":703,"description":704,"org":705,"tags":706,"stars":22,"repoUrl":23,"updatedAt":717},"cross-source-gap-analysis","identify alignment gaps across communication sources","Compare findings across Slack, GitHub, NVIDIA forums, and Outlook to identify alignment gaps, missing coverage, and follow-ups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[707,710,711,714],{"name":708,"slug":709,"type":15},"Analysis","analysis",{"name":569,"slug":570,"type":15},{"name":712,"slug":713,"type":15},"Outlook Email","outlook-email",{"name":715,"slug":716,"type":15},"Slack","slack","2026-08-21T03:36:36.438375",{"slug":719,"name":719,"fn":720,"description":721,"org":722,"tags":723,"stars":22,"repoUrl":23,"updatedAt":731},"deep-research","queue deep multi-step research tasks","Queue deep, multi-step research and analysis tasks to the DeepAgents worker. Supports execution depth presets, request-specific rubrics, SubAgent delegation, and RubricMiddleware cross-validation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[724,725,728],{"name":13,"slug":14,"type":15},{"name":726,"slug":727,"type":15},"Multi-Agent","multi-agent",{"name":729,"slug":730,"type":15},"Research","research","2026-08-31T09:19:11.215563",{"slug":733,"name":733,"fn":734,"description":735,"org":736,"tags":737,"stars":22,"repoUrl":23,"updatedAt":743},"github-readonly-live","read live GitHub repository data","Read an allowed live GitHub repository through authenticated, policy-scoped GitHub REST GET requests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[738,741,742],{"name":739,"slug":740,"type":15},"API Development","api-development",{"name":569,"slug":570,"type":15},{"name":9,"slug":8,"type":15},"2026-08-21T03:36:35.451173",{"slug":745,"name":745,"fn":746,"description":747,"org":748,"tags":749,"stars":22,"repoUrl":23,"updatedAt":759},"gitlab-readonly-live","read GitLab project data","Read one of the configured GitLab projects through authenticated, policy-scoped REST GET requests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[750,753,756],{"name":751,"slug":752,"type":15},"Git","git",{"name":754,"slug":755,"type":15},"GitLab","gitlab",{"name":757,"slug":758,"type":15},"REST API","rest-api","2026-08-31T09:19:36.855848",{"slug":140,"name":140,"fn":761,"description":762,"org":763,"tags":764,"stars":22,"repoUrl":23,"updatedAt":773},"triage incoming email and Slack messages","Decide which newly-arrived email and Slack messages create an obligation for the user, and craft the row that represents each one.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[765,768,771,772],{"name":766,"slug":767,"type":15},"Email","email",{"name":769,"slug":770,"type":15},"Messaging","messaging",{"name":715,"slug":716,"type":15},{"name":583,"slug":584,"type":15},"2026-08-31T09:19:11.560469",28]