[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-braintrust-braintrust-analyze-eval-experiment":3,"mdc--uye9dz-key":38,"related-org-braintrust-braintrust-analyze-eval-experiment":341,"related-repo-braintrust-braintrust-analyze-eval-experiment":508},{"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-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},"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},"LLM","llm","tag",{"name":18,"slug":19,"type":16},"Analysis","analysis",{"name":21,"slug":22,"type":16},"Evals","evals",{"name":24,"slug":25,"type":16},"Statistics","statistics",{"name":9,"slug":8,"type":16},7,"https:\u002F\u002Fgithub.com\u002Fbraintrustdata\u002Feval-library","2026-08-20T03:53:01.13806",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-analyze-eval-experiment","---\nname: braintrust-analyze-eval-experiment\ndescription: >-\n  Analyze completed LLM or agent eval experiments using uncertainty-aware and decision-relevant\n  methods. Use to audit run completeness and pairing, calculate confidence intervals, run paired\n  comparisons, report wins, losses, and ties, incorporate run-to-run variance, handle multiple\n  comparisons, inspect subgroup performance, and test fragility to favorable slices. Use when\n  results already exist and someone asks what they mean, whether a difference is real, or which\n  model won. Do not use to design an experiment that has not yet collected results.\n---\n\n# Analyze completed eval results\n\nContract: `references\u002Finteraction-contract.md`. Calibration, templates, provenance: `references\u002Festimators.md`.\n\n## Trigger\n\n- Completed results needing interpretation: \"analyze these,\" \"is this difference real?\"\n- Requests for intervals, subgroups, multiplicity handling, fragility checks.\n- A headline number about to be quoted with no interval.\n\n## Do\n\n1. **Audit before computing anything.** Reconcile effective N per arm — attempted, completed,\n   errored — against the design, and confirm pairing keys, run counts, and config versions. An arm\n   whose effective N is materially below its peers is not comparable: re-run it rather than\n   caveating it, because the survivors are not a random sample.\n2. Report every headline number as **point estimate + 95% CI + n + K**, using Wilson near 0 or 1.\n3. Apply the two corrections practitioners miss: **cluster standard errors** when items are\n   related, and **fold in run-to-run variance** as a second component.\n4. Compare **paired** on the same items, reporting the paired difference with its CI **plus\n   wins\u002Flosses\u002Fties** — never two independent averages.\n5. Handle multiplicity: record every arm run including failures, then either correct or treat the\n   sweep as exploratory and re-run the winner on a held-out confirmatory set.\n6. Look past the average: run distribution and **worst run**; slices by metadata strata; the\n   breakdown check — does the conclusion survive dropping the most favorable category?\n\n## Avoid\n\n- Do not read any average before the effective-N reconciliation.\n- Do not compare independent averages when item-level pairing is available.\n- Do not promote a sweep winner without confirmation — that is an anecdote with a p-value.\n- Do not fold errored items into the numerator silently, or adopt a scorer fix because it improves\n  one arm.\n- Do not write the external report or make the ship call here. If arms differ in more than one\n  respect, isolate the factors before attributing the effect to any of them.\n\n## Check\n\n- Effective N per arm and how errors were treated; every estimate carries interval, n, K.\n- Clustering and run variance accounted for; paired differences with wins\u002Flosses\u002Fties.\n- Search denominator disclosed; exploratory vs. confirmatory labeled.\n- Subgroup slices and at least one fragility check reported.\n\n## Risk\n\n- Clustering, silent exclusions, repeated benchmark touches, and favorable slices each make\n  uncertainty look far smaller than it is.\n- Two systems can share an average while differing in reliability; the mean hides what users\n  experience.\n- Subgroups can rank oppositely to the aggregate, making the aggregate the least informative\n  number available.\n- Fragility is highest among top-ranked systems, exactly where \"best\" claims concentrate.\n\n## Braintrust\n\nShared mechanics: `references\u002Fplatform-mechanics.md`. Three are load-bearing here, in this\norder. The **read-safety checks (§2) run before any average** — this skill's first step *is*\neffective-N reconciliation, and a truncated pull looks exactly like a missing stratum.\n**Pairing (§3)** decides whether step 4 is available at all. **§8** covers every statistic this\nskill asks for that the platform will not compute: κ, clustered SEs, heatmaps.\n\nSubgroup slices come from grouping by `metadata`. Where a stratum was never written, report the\ngap as an instrumentation finding rather than quietly dropping the slice — it is the one analysis\nlimitation with no retroactive fix.\n",{"data":39,"body":40},{"name":4,"description":6},{"type":41,"children":42},"root",[43,52,75,82,102,108,190,196,224,230,253,259,282,286,328],{"type":44,"tag":45,"props":46,"children":48},"element","h1",{"id":47},"analyze-completed-eval-results",[49],{"type":50,"value":51},"text","Analyze completed eval results",{"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\u002Festimators.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},"Completed results needing interpretation: \"analyze these,\" \"is this difference real?\"",{"type":44,"tag":87,"props":93,"children":94},{},[95],{"type":50,"value":96},"Requests for intervals, subgroups, multiplicity handling, fragility checks.",{"type":44,"tag":87,"props":98,"children":99},{},[100],{"type":50,"value":101},"A headline number about to be quoted with no interval.",{"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",{},[112,123,135,154,173,178],{"type":44,"tag":87,"props":113,"children":114},{},[115,121],{"type":44,"tag":116,"props":117,"children":118},"strong",{},[119],{"type":50,"value":120},"Audit before computing anything.",{"type":50,"value":122}," Reconcile effective N per arm — attempted, completed,\nerrored — against the design, and confirm pairing keys, run counts, and config versions. An arm\nwhose effective N is materially below its peers is not comparable: re-run it rather than\ncaveating it, because the survivors are not a random sample.",{"type":44,"tag":87,"props":124,"children":125},{},[126,128,133],{"type":50,"value":127},"Report every headline number as ",{"type":44,"tag":116,"props":129,"children":130},{},[131],{"type":50,"value":132},"point estimate + 95% CI + n + K",{"type":50,"value":134},", using Wilson near 0 or 1.",{"type":44,"tag":87,"props":136,"children":137},{},[138,140,145,147,152],{"type":50,"value":139},"Apply the two corrections practitioners miss: ",{"type":44,"tag":116,"props":141,"children":142},{},[143],{"type":50,"value":144},"cluster standard errors",{"type":50,"value":146}," when items are\nrelated, and ",{"type":44,"tag":116,"props":148,"children":149},{},[150],{"type":50,"value":151},"fold in run-to-run variance",{"type":50,"value":153}," as a second component.",{"type":44,"tag":87,"props":155,"children":156},{},[157,159,164,166,171],{"type":50,"value":158},"Compare ",{"type":44,"tag":116,"props":160,"children":161},{},[162],{"type":50,"value":163},"paired",{"type":50,"value":165}," on the same items, reporting the paired difference with its CI ",{"type":44,"tag":116,"props":167,"children":168},{},[169],{"type":50,"value":170},"plus\nwins\u002Flosses\u002Fties",{"type":50,"value":172}," — never two independent averages.",{"type":44,"tag":87,"props":174,"children":175},{},[176],{"type":50,"value":177},"Handle multiplicity: record every arm run including failures, then either correct or treat the\nsweep as exploratory and re-run the winner on a held-out confirmatory set.",{"type":44,"tag":87,"props":179,"children":180},{},[181,183,188],{"type":50,"value":182},"Look past the average: run distribution and ",{"type":44,"tag":116,"props":184,"children":185},{},[186],{"type":50,"value":187},"worst run",{"type":50,"value":189},"; slices by metadata strata; the\nbreakdown check — does the conclusion survive dropping the most favorable category?",{"type":44,"tag":76,"props":191,"children":193},{"id":192},"avoid",[194],{"type":50,"value":195},"Avoid",{"type":44,"tag":83,"props":197,"children":198},{},[199,204,209,214,219],{"type":44,"tag":87,"props":200,"children":201},{},[202],{"type":50,"value":203},"Do not read any average before the effective-N reconciliation.",{"type":44,"tag":87,"props":205,"children":206},{},[207],{"type":50,"value":208},"Do not compare independent averages when item-level pairing is available.",{"type":44,"tag":87,"props":210,"children":211},{},[212],{"type":50,"value":213},"Do not promote a sweep winner without confirmation — that is an anecdote with a p-value.",{"type":44,"tag":87,"props":215,"children":216},{},[217],{"type":50,"value":218},"Do not fold errored items into the numerator silently, or adopt a scorer fix because it improves\none arm.",{"type":44,"tag":87,"props":220,"children":221},{},[222],{"type":50,"value":223},"Do not write the external report or make the ship call here. If arms differ in more than one\nrespect, isolate the factors before attributing the effect to any of them.",{"type":44,"tag":76,"props":225,"children":227},{"id":226},"check",[228],{"type":50,"value":229},"Check",{"type":44,"tag":83,"props":231,"children":232},{},[233,238,243,248],{"type":44,"tag":87,"props":234,"children":235},{},[236],{"type":50,"value":237},"Effective N per arm and how errors were treated; every estimate carries interval, n, K.",{"type":44,"tag":87,"props":239,"children":240},{},[241],{"type":50,"value":242},"Clustering and run variance accounted for; paired differences with wins\u002Flosses\u002Fties.",{"type":44,"tag":87,"props":244,"children":245},{},[246],{"type":50,"value":247},"Search denominator disclosed; exploratory vs. confirmatory labeled.",{"type":44,"tag":87,"props":249,"children":250},{},[251],{"type":50,"value":252},"Subgroup slices and at least one fragility check reported.",{"type":44,"tag":76,"props":254,"children":256},{"id":255},"risk",[257],{"type":50,"value":258},"Risk",{"type":44,"tag":83,"props":260,"children":261},{},[262,267,272,277],{"type":44,"tag":87,"props":263,"children":264},{},[265],{"type":50,"value":266},"Clustering, silent exclusions, repeated benchmark touches, and favorable slices each make\nuncertainty look far smaller than it is.",{"type":44,"tag":87,"props":268,"children":269},{},[270],{"type":50,"value":271},"Two systems can share an average while differing in reliability; the mean hides what users\nexperience.",{"type":44,"tag":87,"props":273,"children":274},{},[275],{"type":50,"value":276},"Subgroups can rank oppositely to the aggregate, making the aggregate the least informative\nnumber available.",{"type":44,"tag":87,"props":278,"children":279},{},[280],{"type":50,"value":281},"Fragility is highest among top-ranked systems, exactly where \"best\" claims concentrate.",{"type":44,"tag":76,"props":283,"children":284},{"id":8},[285],{"type":50,"value":9},{"type":44,"tag":53,"props":287,"children":288},{},[289,291,297,299,304,306,312,314,319,321,326],{"type":50,"value":290},"Shared mechanics: ",{"type":44,"tag":59,"props":292,"children":294},{"className":293},[],[295],{"type":50,"value":296},"references\u002Fplatform-mechanics.md",{"type":50,"value":298},". Three are load-bearing here, in this\norder. The ",{"type":44,"tag":116,"props":300,"children":301},{},[302],{"type":50,"value":303},"read-safety checks (§2) run before any average",{"type":50,"value":305}," — this skill's first step ",{"type":44,"tag":307,"props":308,"children":309},"em",{},[310],{"type":50,"value":311},"is",{"type":50,"value":313},"\neffective-N reconciliation, and a truncated pull looks exactly like a missing stratum.\n",{"type":44,"tag":116,"props":315,"children":316},{},[317],{"type":50,"value":318},"Pairing (§3)",{"type":50,"value":320}," decides whether step 4 is available at all. ",{"type":44,"tag":116,"props":322,"children":323},{},[324],{"type":50,"value":325},"§8",{"type":50,"value":327}," covers every statistic this\nskill asks for that the platform will not compute: κ, clustered SEs, heatmaps.",{"type":44,"tag":53,"props":329,"children":330},{},[331,333,339],{"type":50,"value":332},"Subgroup slices come from grouping by ",{"type":44,"tag":59,"props":334,"children":336},{"className":335},[],[337],{"type":50,"value":338},"metadata",{"type":50,"value":340},". Where a stratum was never written, report the\ngap as an instrumentation finding rather than quietly dropping the slice — it is the one analysis\nlimitation with no retroactive fix.",{"items":342,"total":507},[343,359,367,381,397,413,427,439,453,469,481,493],{"slug":344,"name":344,"fn":345,"description":346,"org":347,"tags":348,"stars":356,"repoUrl":357,"updatedAt":358},"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},[349,350,353],{"name":9,"slug":8,"type":16},{"name":351,"slug":352,"type":16},"Debugging","debugging",{"name":354,"slug":355,"type":16},"MCP","mcp",18,"https:\u002F\u002Fgithub.com\u002Fbraintrustdata\u002Fbraintrust-claude-plugin","2026-07-12T08:36:13.889274",{"slug":4,"name":4,"fn":5,"description":6,"org":360,"tags":361,"stars":27,"repoUrl":28,"updatedAt":29},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[362,363,364,365,366],{"name":18,"slug":19,"type":16},{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":24,"slug":25,"type":16},{"slug":368,"name":368,"fn":369,"description":370,"org":371,"tags":372,"stars":27,"repoUrl":28,"updatedAt":380},"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},[373,374,375,376,377],{"name":18,"slug":19,"type":16},{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":378,"slug":379,"type":16},"Performance","performance","2026-08-20T03:53:40.036077",{"slug":382,"name":382,"fn":383,"description":384,"org":385,"tags":386,"stars":27,"repoUrl":28,"updatedAt":396},"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},[387,390,391,394,395],{"name":388,"slug":389,"type":16},"Agents","agents",{"name":9,"slug":8,"type":16},{"name":392,"slug":393,"type":16},"Datasets","datasets",{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},"2026-08-20T03:53:33.304815",{"slug":398,"name":398,"fn":399,"description":400,"org":401,"tags":402,"stars":27,"repoUrl":28,"updatedAt":412},"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},[403,404,405,406,409],{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":407,"slug":408,"type":16},"Product Management","product-management",{"name":410,"slug":411,"type":16},"Strategy","strategy","2026-08-20T03:53:00.07097",{"slug":414,"name":414,"fn":415,"description":416,"org":417,"tags":418,"stars":27,"repoUrl":28,"updatedAt":426},"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},[419,420,421,424,425],{"name":388,"slug":389,"type":16},{"name":9,"slug":8,"type":16},{"name":422,"slug":423,"type":16},"CI\u002FCD","ci-cd",{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},"2026-08-20T03:53:15.686158",{"slug":428,"name":428,"fn":429,"description":430,"org":431,"tags":432,"stars":27,"repoUrl":28,"updatedAt":438},"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},[433,434,437],{"name":9,"slug":8,"type":16},{"name":435,"slug":436,"type":16},"Deployment","deployment",{"name":21,"slug":22,"type":16},"2026-08-20T03:53:32.558937",{"slug":440,"name":440,"fn":441,"description":442,"org":443,"tags":444,"stars":27,"repoUrl":28,"updatedAt":452},"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},[445,446,447,450,451],{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":448,"slug":449,"type":16},"Experiments","experiments",{"name":14,"slug":15,"type":16},{"name":410,"slug":411,"type":16},"2026-08-20T03:53:36.534554",{"slug":454,"name":454,"fn":455,"description":456,"org":457,"tags":458,"stars":27,"repoUrl":28,"updatedAt":468},"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},[459,460,461,462,465],{"name":9,"slug":8,"type":16},{"name":392,"slug":393,"type":16},{"name":21,"slug":22,"type":16},{"name":463,"slug":464,"type":16},"Observability","observability",{"name":466,"slug":467,"type":16},"Tracing","tracing","2026-08-20T03:53:37.274703",{"slug":470,"name":470,"fn":471,"description":472,"org":473,"tags":474,"stars":27,"repoUrl":28,"updatedAt":480},"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},[475,476,477,478,479],{"name":388,"slug":389,"type":16},{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":378,"slug":379,"type":16},"2026-08-20T03:53:16.078609",{"slug":482,"name":482,"fn":483,"description":484,"org":485,"tags":486,"stars":27,"repoUrl":28,"updatedAt":492},"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},[487,488,489,490,491],{"name":388,"slug":389,"type":16},{"name":9,"slug":8,"type":16},{"name":392,"slug":393,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},"2026-08-20T03:53:36.900129",{"slug":494,"name":494,"fn":495,"description":496,"org":497,"tags":498,"stars":27,"repoUrl":28,"updatedAt":506},"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},[499,500,501,502,503],{"name":388,"slug":389,"type":16},{"name":9,"slug":8,"type":16},{"name":351,"slug":352,"type":16},{"name":21,"slug":22,"type":16},{"name":504,"slug":505,"type":16},"Triage","triage","2026-08-20T03:53:00.423941",27,{"items":509,"total":564},[510,518,526,534,542,550,556],{"slug":4,"name":4,"fn":5,"description":6,"org":511,"tags":512,"stars":27,"repoUrl":28,"updatedAt":29},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[513,514,515,516,517],{"name":18,"slug":19,"type":16},{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":24,"slug":25,"type":16},{"slug":368,"name":368,"fn":369,"description":370,"org":519,"tags":520,"stars":27,"repoUrl":28,"updatedAt":380},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[521,522,523,524,525],{"name":18,"slug":19,"type":16},{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":378,"slug":379,"type":16},{"slug":382,"name":382,"fn":383,"description":384,"org":527,"tags":528,"stars":27,"repoUrl":28,"updatedAt":396},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[529,530,531,532,533],{"name":388,"slug":389,"type":16},{"name":9,"slug":8,"type":16},{"name":392,"slug":393,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"slug":398,"name":398,"fn":399,"description":400,"org":535,"tags":536,"stars":27,"repoUrl":28,"updatedAt":412},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[537,538,539,540,541],{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"name":407,"slug":408,"type":16},{"name":410,"slug":411,"type":16},{"slug":414,"name":414,"fn":415,"description":416,"org":543,"tags":544,"stars":27,"repoUrl":28,"updatedAt":426},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[545,546,547,548,549],{"name":388,"slug":389,"type":16},{"name":9,"slug":8,"type":16},{"name":422,"slug":423,"type":16},{"name":21,"slug":22,"type":16},{"name":14,"slug":15,"type":16},{"slug":428,"name":428,"fn":429,"description":430,"org":551,"tags":552,"stars":27,"repoUrl":28,"updatedAt":438},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[553,554,555],{"name":9,"slug":8,"type":16},{"name":435,"slug":436,"type":16},{"name":21,"slug":22,"type":16},{"slug":440,"name":440,"fn":441,"description":442,"org":557,"tags":558,"stars":27,"repoUrl":28,"updatedAt":452},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[559,560,561,562,563],{"name":9,"slug":8,"type":16},{"name":21,"slug":22,"type":16},{"name":448,"slug":449,"type":16},{"name":14,"slug":15,"type":16},{"name":410,"slug":411,"type":16},24]