[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-braintrust-braintrust-attribute-multi-variable-change":3,"mdc--20zi0s-key":38,"related-org-braintrust-braintrust-attribute-multi-variable-change":332,"related-repo-braintrust-braintrust-attribute-multi-variable-change":499},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":12,"stars":27,"repoUrl":28,"updatedAt":29,"license":30,"forks":31,"topics":32,"repo":33,"sourceUrl":36,"mdContent":37},"braintrust-attribute-multi-variable-change","attribute performance changes to multiple variables","Attribute an observed change when several things moved at once — model plus prompt plus tools, a provider migration, a framework upgrade, or a vendor swap that bundles serving stack with model. Use when asked which part of a change caused the result, when a comparison's arms differ in more than one way, when a treatment has no uniform implementation across vendors, or when a serving-stack difference is confounded with a model difference. Do not use for a clean single-variable comparison, or to design an experiment that has not yet run.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},"braintrust","Braintrust","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fbraintrust.png","braintrustdata",[13,17,20,23,26],{"name":14,"slug":15,"type":16},"Performance","performance","tag",{"name":18,"slug":19,"type":16},"LLM","llm",{"name":21,"slug":22,"type":16},"Analysis","analysis",{"name":24,"slug":25,"type":16},"Evals","evals",{"name":9,"slug":8,"type":16},7,"https:\u002F\u002Fgithub.com\u002Fbraintrustdata\u002Feval-library","2026-08-20T03:53:40.036077",null,0,[],{"repoUrl":28,"stars":27,"forks":31,"topics":34,"description":35},[],"Braintrust eval skills library","https:\u002F\u002Fgithub.com\u002Fbraintrustdata\u002Feval-library\u002Ftree\u002FHEAD\u002Fskills\u002Fbraintrust-attribute-multi-variable-change","---\nname: braintrust-attribute-multi-variable-change\ndescription: >-\n  Attribute an observed change when several things moved at once — model plus prompt plus tools, a\n  provider migration, a framework upgrade, or a vendor swap that bundles serving stack with model.\n  Use when asked which part of a change caused the result, when a comparison's arms differ in more\n  than one way, when a treatment has no uniform implementation across vendors, or when a\n  serving-stack difference is confounded with a model difference. Do not use for a clean\n  single-variable comparison, or to design an experiment that has not yet run.\n---\n\n# Attribute a multi-variable change\n\nContract: `references\u002Finteraction-contract.md`. Calibration, templates, provenance: `references\u002Fisolation-designs.md`.\n\n## Trigger\n\n- \"We upgraded the model and rewrote the prompt — which one helped?\"\n- A vendor or provider migration bundling model, serving stack, and tool surface.\n- Arms that differ in more than one respect, discovered after results exist.\n\n## Do\n\n1. Enumerate what actually differs between arms with the inventory in `reference.md`, including\n   the variables hiding inside apparently atomic ones: the serving path behind a model string, the\n   tool manifest behind \"the agent.\"\n2. State honestly what the existing data can and cannot separate. Two arms differing in three ways\n   support **one** claim about the bundle and **no** claim about any component. Say that before\n   analyzing anything.\n3. Choose the cheapest design that isolates what the decision needs — ablation ladder,\n   one-factor-at-a-time from the new baseline, or full factorial only when interactions are the\n   actual question.\n4. Where a factor cannot be held constant — a vendor with no equivalent parameter — record the\n   per-arm implementation as a published table and **exclude that arm from the component claim**\n   rather than scoring it \"no benefit.\"\n5. Report bundle effect and component effects as **separate rows**, each with its own uncertainty.\n\n## Avoid\n\n- Do not attribute a bundle's gain to its most interesting component because it is the one you\n  care about.\n- Do not treat a provider swap as a model comparison; precision, batching, and hardware shift\n  outputs independent of weights.\n- Do not fix arms mid-comparison to make them match — restart, or finish and document the\n  confound.\n- Do not reach for this on a clean single-variable comparison; ordinary paired analysis is\n  both simpler and more sensitive.\n\n## Check\n\n- Full difference inventory per arm, including serving path and tool manifest.\n- An explicit statement of which claims the current data cannot support.\n- Isolation design chosen with its cost, and the factor each run isolates.\n- Per-arm implementation table for any non-uniform treatment.\n- Bundle and component effects reported separately, never merged.\n\n## Risk\n\n- The most common outcome is that the data supports no component claim at all, and **saying so is\n  the correct deliverable** — an attribution invented from a confounded comparison is worse than\n  none.\n- Interactions are real: two changes can each help alone and hurt together, so\n  one-factor-at-a-time results do not simply add.\n- Serving-stack variation is the confound teams most often leave unnamed, because it rides inside\n  a model string that looks like a single value.\n\n## Braintrust\n\nShared mechanics: `references\u002Fplatform-mechanics.md`. Attribution leans hardest on **§4\nmetadata** — the per-arm record of model string, decoding params, provider\u002Fendpoint\u002Ftier, cache\nstate, and prompt\u002Fscorer\u002Ftool versions *is* the difference inventory, recoverable months later.\nWithout it, attribution is guesswork. **§7 hygiene** matters more here than elsewhere too, since\na throttled provider left unthrottled becomes one more confound in a comparison already carrying\nseveral.\n\nName for the factor being toggled, **most-significant first**, so an ablation ladder reads down\nthe experiment list in order.\n\nUse cross-experiment diffs for item-paired component estimates, and group by `metadata` to check\nwhether a component effect is concentrated in one stratum; a component helping only one slice is\na stratification finding, not a main effect.\n",{"data":39,"body":40},{"name":4,"description":6},{"type":41,"children":42},"root",[43,52,75,82,102,108,174,180,203,209,237,243,268,272,307,319],{"type":44,"tag":45,"props":46,"children":48},"element","h1",{"id":47},"attribute-a-multi-variable-change",[49],{"type":50,"value":51},"text","Attribute a multi-variable change",{"type":44,"tag":53,"props":54,"children":55},"p",{},[56,58,65,67,73],{"type":50,"value":57},"Contract: ",{"type":44,"tag":59,"props":60,"children":62},"code",{"className":61},[],[63],{"type":50,"value":64},"references\u002Finteraction-contract.md",{"type":50,"value":66},". Calibration, templates, provenance: ",{"type":44,"tag":59,"props":68,"children":70},{"className":69},[],[71],{"type":50,"value":72},"references\u002Fisolation-designs.md",{"type":50,"value":74},".",{"type":44,"tag":76,"props":77,"children":79},"h2",{"id":78},"trigger",[80],{"type":50,"value":81},"Trigger",{"type":44,"tag":83,"props":84,"children":85},"ul",{},[86,92,97],{"type":44,"tag":87,"props":88,"children":89},"li",{},[90],{"type":50,"value":91},"\"We upgraded the model and rewrote the prompt — which one helped?\"",{"type":44,"tag":87,"props":93,"children":94},{},[95],{"type":50,"value":96},"A vendor or provider migration bundling model, serving stack, and tool surface.",{"type":44,"tag":87,"props":98,"children":99},{},[100],{"type":50,"value":101},"Arms that differ in more than one respect, discovered after results exist.",{"type":44,"tag":76,"props":103,"children":105},{"id":104},"do",[106],{"type":50,"value":107},"Do",{"type":44,"tag":109,"props":110,"children":111},"ol",{},[112,125,145,150,162],{"type":44,"tag":87,"props":113,"children":114},{},[115,117,123],{"type":50,"value":116},"Enumerate what actually differs between arms with the inventory in ",{"type":44,"tag":59,"props":118,"children":120},{"className":119},[],[121],{"type":50,"value":122},"reference.md",{"type":50,"value":124},", including\nthe variables hiding inside apparently atomic ones: the serving path behind a model string, the\ntool manifest behind \"the agent.\"",{"type":44,"tag":87,"props":126,"children":127},{},[128,130,136,138,143],{"type":50,"value":129},"State honestly what the existing data can and cannot separate. Two arms differing in three ways\nsupport ",{"type":44,"tag":131,"props":132,"children":133},"strong",{},[134],{"type":50,"value":135},"one",{"type":50,"value":137}," claim about the bundle and ",{"type":44,"tag":131,"props":139,"children":140},{},[141],{"type":50,"value":142},"no",{"type":50,"value":144}," claim about any component. Say that before\nanalyzing anything.",{"type":44,"tag":87,"props":146,"children":147},{},[148],{"type":50,"value":149},"Choose the cheapest design that isolates what the decision needs — ablation ladder,\none-factor-at-a-time from the new baseline, or full factorial only when interactions are the\nactual question.",{"type":44,"tag":87,"props":151,"children":152},{},[153,155,160],{"type":50,"value":154},"Where a factor cannot be held constant — a vendor with no equivalent parameter — record the\nper-arm implementation as a published table and ",{"type":44,"tag":131,"props":156,"children":157},{},[158],{"type":50,"value":159},"exclude that arm from the component claim",{"type":50,"value":161},"\nrather than scoring it \"no benefit.\"",{"type":44,"tag":87,"props":163,"children":164},{},[165,167,172],{"type":50,"value":166},"Report bundle effect and component effects as ",{"type":44,"tag":131,"props":168,"children":169},{},[170],{"type":50,"value":171},"separate rows",{"type":50,"value":173},", each with its own uncertainty.",{"type":44,"tag":76,"props":175,"children":177},{"id":176},"avoid",[178],{"type":50,"value":179},"Avoid",{"type":44,"tag":83,"props":181,"children":182},{},[183,188,193,198],{"type":44,"tag":87,"props":184,"children":185},{},[186],{"type":50,"value":187},"Do not attribute a bundle's gain to its most interesting component because it is the one you\ncare about.",{"type":44,"tag":87,"props":189,"children":190},{},[191],{"type":50,"value":192},"Do not treat a provider swap as a model comparison; precision, batching, and hardware shift\noutputs independent of weights.",{"type":44,"tag":87,"props":194,"children":195},{},[196],{"type":50,"value":197},"Do not fix arms mid-comparison to make them match — restart, or finish and document the\nconfound.",{"type":44,"tag":87,"props":199,"children":200},{},[201],{"type":50,"value":202},"Do not reach for this on a clean single-variable comparison; ordinary paired analysis is\nboth simpler and more sensitive.",{"type":44,"tag":76,"props":204,"children":206},{"id":205},"check",[207],{"type":50,"value":208},"Check",{"type":44,"tag":83,"props":210,"children":211},{},[212,217,222,227,232],{"type":44,"tag":87,"props":213,"children":214},{},[215],{"type":50,"value":216},"Full difference inventory per arm, including serving path and tool manifest.",{"type":44,"tag":87,"props":218,"children":219},{},[220],{"type":50,"value":221},"An explicit statement of which claims the current data cannot support.",{"type":44,"tag":87,"props":223,"children":224},{},[225],{"type":50,"value":226},"Isolation design chosen with its cost, and the factor each run isolates.",{"type":44,"tag":87,"props":228,"children":229},{},[230],{"type":50,"value":231},"Per-arm implementation table for any non-uniform treatment.",{"type":44,"tag":87,"props":233,"children":234},{},[235],{"type":50,"value":236},"Bundle and component effects reported separately, never merged.",{"type":44,"tag":76,"props":238,"children":240},{"id":239},"risk",[241],{"type":50,"value":242},"Risk",{"type":44,"tag":83,"props":244,"children":245},{},[246,258,263],{"type":44,"tag":87,"props":247,"children":248},{},[249,251,256],{"type":50,"value":250},"The most common outcome is that the data supports no component claim at all, and ",{"type":44,"tag":131,"props":252,"children":253},{},[254],{"type":50,"value":255},"saying so is\nthe correct deliverable",{"type":50,"value":257}," — an attribution invented from a confounded comparison is worse than\nnone.",{"type":44,"tag":87,"props":259,"children":260},{},[261],{"type":50,"value":262},"Interactions are real: two changes can each help alone and hurt together, so\none-factor-at-a-time results do not simply add.",{"type":44,"tag":87,"props":264,"children":265},{},[266],{"type":50,"value":267},"Serving-stack variation is the confound teams most often leave unnamed, because it rides inside\na model string that looks like a single value.",{"type":44,"tag":76,"props":269,"children":270},{"id":8},[271],{"type":50,"value":9},{"type":44,"tag":53,"props":273,"children":274},{},[275,277,283,285,290,292,298,300,305],{"type":50,"value":276},"Shared mechanics: ",{"type":44,"tag":59,"props":278,"children":280},{"className":279},[],[281],{"type":50,"value":282},"references\u002Fplatform-mechanics.md",{"type":50,"value":284},". Attribution leans hardest on ",{"type":44,"tag":131,"props":286,"children":287},{},[288],{"type":50,"value":289},"§4\nmetadata",{"type":50,"value":291}," — the per-arm record of model string, decoding params, provider\u002Fendpoint\u002Ftier, cache\nstate, and prompt\u002Fscorer\u002Ftool versions ",{"type":44,"tag":293,"props":294,"children":295},"em",{},[296],{"type":50,"value":297},"is",{"type":50,"value":299}," the difference inventory, recoverable months later.\nWithout it, attribution is guesswork. ",{"type":44,"tag":131,"props":301,"children":302},{},[303],{"type":50,"value":304},"§7 hygiene",{"type":50,"value":306}," matters more here than elsewhere too, since\na throttled provider left unthrottled becomes one more confound in a comparison already carrying\nseveral.",{"type":44,"tag":53,"props":308,"children":309},{},[310,312,317],{"type":50,"value":311},"Name for the factor being toggled, ",{"type":44,"tag":131,"props":313,"children":314},{},[315],{"type":50,"value":316},"most-significant first",{"type":50,"value":318},", so an ablation ladder reads down\nthe experiment list in order.",{"type":44,"tag":53,"props":320,"children":321},{},[322,324,330],{"type":50,"value":323},"Use cross-experiment diffs for item-paired component estimates, and group by ",{"type":44,"tag":59,"props":325,"children":327},{"className":326},[],[328],{"type":50,"value":329},"metadata",{"type":50,"value":331}," to check\nwhether a component effect is concentrated in one stratum; a component helping only one slice is\na stratification finding, not a main effect.",{"items":333,"total":498},[334,350,364,372,388,404,418,430,444,460,472,484],{"slug":335,"name":335,"fn":336,"description":337,"org":338,"tags":339,"stars":347,"repoUrl":348,"updatedAt":349},"troubleshoot-braintrust-mcp","configure and troubleshoot Braintrust MCP servers","This plugin auto-configures a \"braintrust\" MCP server. If you can't see it or reach it, activate this skill\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[340,341,344],{"name":9,"slug":8,"type":16},{"name":342,"slug":343,"type":16},"Debugging","debugging",{"name":345,"slug":346,"type":16},"MCP","mcp",18,"https:\u002F\u002Fgithub.com\u002Fbraintrustdata\u002Fbraintrust-claude-plugin","2026-07-12T08:36:13.889274",{"slug":351,"name":351,"fn":352,"description":353,"org":354,"tags":355,"stars":27,"repoUrl":28,"updatedAt":363},"braintrust-analyze-eval-experiment","analyze LLM and agent eval experiments","Analyze completed LLM or agent eval experiments using uncertainty-aware and decision-relevant methods. Use to audit run completeness and pairing, calculate confidence intervals, run paired comparisons, report wins, losses, and ties, incorporate run-to-run variance, handle multiple comparisons, inspect subgroup performance, and test fragility to favorable slices. Use when results already exist and someone asks what they mean, whether a difference is real, or which model won. Do not use to design an experiment that has not yet collected results.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[356,357,358,359,360],{"name":21,"slug":22,"type":16},{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":361,"slug":362,"type":16},"Statistics","statistics","2026-08-20T03:53:01.13806",{"slug":4,"name":4,"fn":5,"description":6,"org":365,"tags":366,"stars":27,"repoUrl":28,"updatedAt":29},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[367,368,369,370,371],{"name":21,"slug":22,"type":16},{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},{"slug":373,"name":373,"fn":374,"description":375,"org":376,"tags":377,"stars":27,"repoUrl":28,"updatedAt":387},"braintrust-build-eval-dataset","create and manage LLM eval datasets","Create, edit, audit, or compare eval datasets for LLM applications and agents, including target-population definition, case sourcing from production traces, stratified sampling, label provenance and label audits, expected values as constraints for open-ended tasks, dev\u002Ftest splits, contamination and leakage controls, headroom checks, refresh policy, and datasheets. Use when working on the content or lifecycle of an eval dataset. Do not use for sample-size or power calculations, scorer implementation, or open-ended adversarial failure discovery.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[378,381,382,385,386],{"name":379,"slug":380,"type":16},"Agents","agents",{"name":9,"slug":8,"type":16},{"name":383,"slug":384,"type":16},"Datasets","datasets",{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},"2026-08-20T03:53:33.304815",{"slug":389,"name":389,"fn":390,"description":391,"org":392,"tags":393,"stars":27,"repoUrl":28,"updatedAt":403},"braintrust-define-eval-objective","define LLM evaluation objectives","Create, edit, or audit an eval objective by working backward from a product decision to the target outcome, construct, population, intended claim, and verification-versus-validation questions. Use when a team is unsure what an eval should establish, asks \"what are we actually trying to measure,\" \"is this eval measuring the right thing,\" \"does this benchmark support our claim,\" or needs to turn a product goal into an eval objective and state which claims are out of scope. Do not use to select detailed metrics, design datasets, or implement scorers.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[394,395,396,397,400],{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":398,"slug":399,"type":16},"Product Management","product-management",{"name":401,"slug":402,"type":16},"Strategy","strategy","2026-08-20T03:53:00.07097",{"slug":405,"name":405,"fn":406,"description":407,"org":408,"tags":409,"stars":27,"repoUrl":28,"updatedAt":417},"braintrust-define-eval-release-gate","configure release gates for LLM applications","Create, edit, audit, or apply release gates for LLM applications and agents. Use to combine minimum meaningful improvement, statistical significance, regression rate, subgroup consistency, worst-run stability, all-attempts reliability, safety upper bounds, latency, and cost into an explicit ship-or-hold policy, to turn metrics into a CI gate, or to explain why a candidate failed a gate and what evidence would justify reconsideration. Do not use for general result analysis without a deployment decision.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[410,411,412,415,416],{"name":379,"slug":380,"type":16},{"name":9,"slug":8,"type":16},{"name":413,"slug":414,"type":16},"CI\u002FCD","ci-cd",{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},"2026-08-20T03:53:15.686158",{"slug":419,"name":419,"fn":420,"description":421,"org":422,"tags":423,"stars":27,"repoUrl":28,"updatedAt":429},"braintrust-deploy-evaluator","deploy evaluators to Braintrust","Take a validated scorer or classifier from definition to running instrument in Braintrust — scope selection, inline testing before saving, saving as an evaluator, attaching an online-scoring rule, activating it for new traffic, and backfilling history with a rewind. Use when a scorer needs to actually run against production logs, when an online-scoring rule needs to be created or changed, or when historical traces need scoring. Do not use to decide what the scorer should measure, to write its rubric, or to establish that it agrees with human judgment.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[424,425,428],{"name":9,"slug":8,"type":16},{"name":426,"slug":427,"type":16},"Deployment","deployment",{"name":24,"slug":25,"type":16},"2026-08-20T03:53:32.558937",{"slug":431,"name":431,"fn":432,"description":433,"org":434,"tags":435,"stars":27,"repoUrl":28,"updatedAt":443},"braintrust-design-eval-experiment","design controlled LLM eval experiments","Design or audit controlled eval experiments for model, prompt, retrieval, tool, guardrail, or agent-architecture changes. Use before data collection to state directional and minimum-effect hypotheses, name independent, dependent, and control variables including the serving environment and tool surface, choose paired designs, set repetitions and allocation, distinguish exploratory from confirmatory comparisons, and pre-specify stopping, exclusion, multiplicity, and analysis rules. Do not use primarily to analyze results already collected.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[436,437,438,441,442],{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":439,"slug":440,"type":16},"Experiments","experiments",{"name":18,"slug":19,"type":16},{"name":401,"slug":402,"type":16},"2026-08-20T03:53:36.534554",{"slug":445,"name":445,"fn":446,"description":447,"org":448,"tags":449,"stars":27,"repoUrl":28,"updatedAt":459},"braintrust-design-eval-instrumentation","design trace and evaluation dataset schemas","Design the trace and eval-dataset schema for an LLM app or agent, and wire the system to emit it. Use when deciding what to log, designing a trace schema, setting up tracing or observability before evals, or when failures cannot be debugged or sliced from existing traces — covering inputs, outputs, spans for tool and LLM calls, state changes, metadata, resolved configuration, serving path, tool manifest, per-item status, attachments, and subgroup variables. Do not use to decide what the evidence should mean, or to write scorers.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[450,451,452,453,456],{"name":9,"slug":8,"type":16},{"name":383,"slug":384,"type":16},{"name":24,"slug":25,"type":16},{"name":454,"slug":455,"type":16},"Observability","observability",{"name":457,"slug":458,"type":16},"Tracing","tracing","2026-08-20T03:53:37.274703",{"slug":461,"name":461,"fn":462,"description":463,"org":464,"tags":465,"stars":27,"repoUrl":28,"updatedAt":471},"braintrust-design-eval-metric-bundle","create multi-objective evaluation metric bundles","Create, edit, audit, or compare a multi-objective eval metric bundle covering product quality, safety, reliability, latency, and cost. Use to choose metrics for an eval, define a goodness bundle, distinguish optimization metrics from non-regression guardrails, expose tradeoffs, audit a KPI or single composite score for Goodhart and metric-gaming risk, or answer \"what should improve and what must not regress.\" Do not use to design trace schemas, build datasets, or implement scoring methods.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[466,467,468,469,470],{"name":379,"slug":380,"type":16},{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},"2026-08-20T03:53:16.078609",{"slug":473,"name":473,"fn":474,"description":475,"org":476,"tags":477,"stars":27,"repoUrl":28,"updatedAt":483},"braintrust-design-human-eval-review","design human evaluation and review workflows","Design or audit human evaluation workflows and golden datasets for LLM applications and agents. Use to set up expert review, select review cases, write reviewer instructions, assign raters, capture rationales and confidence, measure inter-rater agreement with kappa or alpha, adjudicate disagreements, and preserve reviewed examples with provenance as a versioned reference set. Do not use to elicit the criteria or rubric in the first place, to validate a scorer once reference labels exist, or to implement the scorer.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[478,479,480,481,482],{"name":379,"slug":380,"type":16},{"name":9,"slug":8,"type":16},{"name":383,"slug":384,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},"2026-08-20T03:53:36.900129",{"slug":485,"name":485,"fn":486,"description":487,"org":488,"tags":489,"stars":27,"repoUrl":28,"updatedAt":497},"braintrust-discover-agent-failures","identify and classify agent failure modes","Search open-endedly for unanticipated agent failure modes and convert them into a named taxonomy and durable regression items. Use for requests to find out what goes wrong, surface unknown or silent failures, do error analysis over traces, cluster and triage production failures, or build a failure taxonomy — where the goal is discovering modes nobody thought to test rather than measuring a predefined criterion. Produces datasets and taxonomies, not headline scores. Do not use for adversarial attacks against a threat model, or to measure a known criterion.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[490,491,492,493,494],{"name":379,"slug":380,"type":16},{"name":9,"slug":8,"type":16},{"name":342,"slug":343,"type":16},{"name":24,"slug":25,"type":16},{"name":495,"slug":496,"type":16},"Triage","triage","2026-08-20T03:53:00.423941",27,{"items":500,"total":555},[501,509,517,525,533,541,547],{"slug":351,"name":351,"fn":352,"description":353,"org":502,"tags":503,"stars":27,"repoUrl":28,"updatedAt":363},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[504,505,506,507,508],{"name":21,"slug":22,"type":16},{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":361,"slug":362,"type":16},{"slug":4,"name":4,"fn":5,"description":6,"org":510,"tags":511,"stars":27,"repoUrl":28,"updatedAt":29},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[512,513,514,515,516],{"name":21,"slug":22,"type":16},{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},{"slug":373,"name":373,"fn":374,"description":375,"org":518,"tags":519,"stars":27,"repoUrl":28,"updatedAt":387},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[520,521,522,523,524],{"name":379,"slug":380,"type":16},{"name":9,"slug":8,"type":16},{"name":383,"slug":384,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"slug":389,"name":389,"fn":390,"description":391,"org":526,"tags":527,"stars":27,"repoUrl":28,"updatedAt":403},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[528,529,530,531,532],{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"name":398,"slug":399,"type":16},{"name":401,"slug":402,"type":16},{"slug":405,"name":405,"fn":406,"description":407,"org":534,"tags":535,"stars":27,"repoUrl":28,"updatedAt":417},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[536,537,538,539,540],{"name":379,"slug":380,"type":16},{"name":9,"slug":8,"type":16},{"name":413,"slug":414,"type":16},{"name":24,"slug":25,"type":16},{"name":18,"slug":19,"type":16},{"slug":419,"name":419,"fn":420,"description":421,"org":542,"tags":543,"stars":27,"repoUrl":28,"updatedAt":429},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[544,545,546],{"name":9,"slug":8,"type":16},{"name":426,"slug":427,"type":16},{"name":24,"slug":25,"type":16},{"slug":431,"name":431,"fn":432,"description":433,"org":548,"tags":549,"stars":27,"repoUrl":28,"updatedAt":443},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[550,551,552,553,554],{"name":9,"slug":8,"type":16},{"name":24,"slug":25,"type":16},{"name":439,"slug":440,"type":16},{"name":18,"slug":19,"type":16},{"name":401,"slug":402,"type":16},24]