[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-arize-debug-trace":3,"mdc-kz685t-key":50,"related-repo-arize-debug-trace":462,"related-org-arize-debug-trace":544},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":12,"stars":23,"repoUrl":24,"updatedAt":25,"license":26,"forks":27,"topics":28,"repo":45,"sourceUrl":48,"mdContent":49},"debug-trace","diagnose failures using trace investigation","Diagnose failure modes by systematically investigating traces. Trigger when the user explicitly asks for cross-trace diagnosis: \"what's going wrong?\", \"were there errors?\", \"debug this\", \"where is my agent struggling?\". Do NOT trigger on: (1) advice questions (\"what should I do?\"), (2) statistical questions (\"what's the average latency?\"), (3) summarize requests, (4) trace filtering (\"show me traces with errors\"), (5) vague questions (\"is there a problem?\"), (6) unrelated requests.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},"arize","Arize AI","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Farize.jpg","Arize-ai",[13,17,20],{"name":14,"slug":15,"type":16},"Tracing","tracing","tag",{"name":18,"slug":19,"type":16},"Observability","observability",{"name":21,"slug":22,"type":16},"Debugging","debugging",10513,"https:\u002F\u002Fgithub.com\u002FArize-ai\u002Fphoenix","2026-07-12T08:08:10.44243",null,977,[29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44],"agents","ai-monitoring","ai-observability","aiengineering","anthropic","datasets","evals","langchain","llamaindex","llm-eval","llm-evaluation","llmops","llms","openai","prompt-engineering","smolagents",{"repoUrl":24,"stars":23,"forks":27,"topics":46,"description":47},[29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44],"AI Observability & Evaluation","https:\u002F\u002Fgithub.com\u002FArize-ai\u002Fphoenix\u002Ftree\u002FHEAD\u002Fsrc\u002Fphoenix\u002Fserver\u002Fagents\u002Fprompts\u002Fskills\u002Fdebug-trace","---\nname: debug-trace\ndescription: >\n  Diagnose failure modes by systematically investigating traces. Trigger when the user explicitly asks for cross-trace diagnosis: \"what's going wrong?\", \"were there errors?\", \"debug this\", \"where is my agent struggling?\". Do NOT trigger on: (1) advice questions (\"what should I do?\"), (2) statistical questions (\"what's the average latency?\"), (3) summarize requests, (4) trace filtering (\"show me traces with errors\"), (5) vague questions (\"is there a problem?\"), (6) unrelated requests.\nsummary: Investigate traces to identify concrete failure modes, likely root causes, and prioritized fixes.\n---\n\n### Orientation\n\nYour goal is to identify common failure modes and provide prioritized, actionable recommendations. If you already have a specific trace in context, skip the Steps below — apply the failure mode checklist directly to that trace and report findings. Otherwise, build a representative picture across multiple traces: start broad, inspect selectively, and stop once the main issue categories are clear.\n\n### Common Failure Modes to Watch For\n\nThis list is non-exhaustive. Use it as a starting checklist, not a complete taxonomy.\n\n- **Explicit errors** — exceptions, error status codes, or error messages in tool call spans, LLM spans, or retriever spans\n- **Cost and latency** — unusually high token counts or slow spans that suggest room for efficiency improvements\n- **Retrieval quality** — irrelevant, missing, or low-scoring chunks in RAG applications\n- **LLM response quality** — hallucination, factual incorrectness, wrong tone, inappropriate refusals\n- **Tool use problems** — wrong tool selected, malformed invocation, or poor handling of the tool response\n- **Trajectory problems** — the agent took an inefficient path, got stuck in a loop, or failed to complete its task\n- **Instrumentation gaps** — the application is poorly traced, leaving visibility gaps that make it difficult to understand behavior\n\n### Steps\n\nDefault query budget: 1–3 orientation queries, 3–5 aggregate or sampling queries, 3–7 targeted drilldowns. Avoid more than 20 GraphQL calls without summarizing progress. Avoid reading more than 5–10 full span inputs or outputs. Prefer fewer, richer queries — use GraphQL aliases to batch independent lookups. Use available GraphQL recipes before writing new queries.\n\n1. **Orient** — Read `\u002Fphoenix\u002Fagent-start.md`. Use the `phoenix-gql` CLI to get a compact project overview in one aliased query: trace count, span count, latency quantiles, token totals, annotation names, slow traces, high-token traces, error spans.\n\n2. **Select** — Choose a representative sample. Prioritize slow traces, high-token traces, and errored traces; include a few normal traces for comparison. Prefer diversity over volume.\n\n3. **Open-code** — For each trace, write free-form notes on problems, surprises, and incorrect behaviors. Focus on the first failure in a trace, since upstream errors often cause downstream issues. Note independent downstream failures only when they reveal a separate root cause.\n\n4. **Axial-code** — Cluster your notes into named failure categories. Let categories emerge from the data. Distinguish exact counts from sampled or estimated counts.\n\n5. **Summarize** — Report findings using the output format below.\n\n### Observation Journal\n\nOutput this table inline in the conversation as you work — it stays compact by design and needs to be visible for axial-coding. Only write it to a file if the user asks for exhaustive analysis across a large number of traces.\n\nFor each inspected trace, add a row. Keep `observations` free-form. Fill in `tentative_category` only when the pattern is clear — leave it blank otherwise.\n\n| trace_id | observations | tentative_category |\n| -------- | ------------ | ------------------ |\n\nUse this journal as the input to axial-coding.\n\n### Output Format\n\n- **Analysis scope** — brief summary of what was analyzed: number of traces examined, time range if relevant, any filters applied\n- **Findings table** — one row per issue category with: label, short description, occurrence count, one or two representative span (or trace) links — see Linking to Findings below\n- **Recommendations** — for each issue, a concrete suggested fix (prompt change, parameter adjustment, tool fix, instrumentation improvement, etc.), if one can be identified\n\n### Linking to Findings\n\nPrefer linking to the specific span that exhibits the issue over the parent trace whenever possible — span links land the user on the exact node, while a trace link forces them to hunt for the relevant span. Fall back to a trace link only when no single span captures the issue (e.g., a trajectory problem spanning many spans).\n\nUse Phoenix's root-relative redirect URLs with the OpenTelemetry IDs returned by GraphQL — no project lookup required. Read the hex OTel IDs from `Span.spanId` and `Trace.traceId`, **not** the `id` field (which is a Relay node ID and will not resolve). A `Span` has no `traceId` field — read it via the nested `trace { traceId }`:\n\n- Span: `[short description](\u002Fredirects\u002Fspans\u002F\u003CspanId>)`\n- Trace: `[short description](\u002Fredirects\u002Ftraces\u002F\u003CtraceId>)`\n\n### Caveats and Pitfalls\n\n- Existing evals and annotations are useful signal, but treat them as one input among many — they may be incomplete or incorrect.\n- Span status codes are not a reliable proxy for whether an error actually occurred. An exception in the output may be expected behavior; a success status code may mask an error message buried in the span attributes.\n- Do not overload your context by reading too many full spans. Read enough to support meaningful recommendations, then stop.\n",{"data":51,"body":53},{"name":4,"description":6,"summary":52},"Investigate traces to identify concrete failure modes, likely root causes, and prioritized fixes.",{"type":54,"children":55},"root",[56,65,71,77,82,158,164,169,240,246,251,272,298,303,309,342,348,353,413,438,444],{"type":57,"tag":58,"props":59,"children":61},"element","h3",{"id":60},"orientation",[62],{"type":63,"value":64},"text","Orientation",{"type":57,"tag":66,"props":67,"children":68},"p",{},[69],{"type":63,"value":70},"Your goal is to identify common failure modes and provide prioritized, actionable recommendations. If you already have a specific trace in context, skip the Steps below — apply the failure mode checklist directly to that trace and report findings. Otherwise, build a representative picture across multiple traces: start broad, inspect selectively, and stop once the main issue categories are clear.",{"type":57,"tag":58,"props":72,"children":74},{"id":73},"common-failure-modes-to-watch-for",[75],{"type":63,"value":76},"Common Failure Modes to Watch For",{"type":57,"tag":66,"props":78,"children":79},{},[80],{"type":63,"value":81},"This list is non-exhaustive. Use it as a starting checklist, not a complete taxonomy.",{"type":57,"tag":83,"props":84,"children":85},"ul",{},[86,98,108,118,128,138,148],{"type":57,"tag":87,"props":88,"children":89},"li",{},[90,96],{"type":57,"tag":91,"props":92,"children":93},"strong",{},[94],{"type":63,"value":95},"Explicit errors",{"type":63,"value":97}," — exceptions, error status codes, or error messages in tool call spans, LLM spans, or retriever spans",{"type":57,"tag":87,"props":99,"children":100},{},[101,106],{"type":57,"tag":91,"props":102,"children":103},{},[104],{"type":63,"value":105},"Cost and latency",{"type":63,"value":107}," — unusually high token counts or slow spans that suggest room for efficiency improvements",{"type":57,"tag":87,"props":109,"children":110},{},[111,116],{"type":57,"tag":91,"props":112,"children":113},{},[114],{"type":63,"value":115},"Retrieval quality",{"type":63,"value":117}," — irrelevant, missing, or low-scoring chunks in RAG applications",{"type":57,"tag":87,"props":119,"children":120},{},[121,126],{"type":57,"tag":91,"props":122,"children":123},{},[124],{"type":63,"value":125},"LLM response quality",{"type":63,"value":127}," — hallucination, factual incorrectness, wrong tone, inappropriate refusals",{"type":57,"tag":87,"props":129,"children":130},{},[131,136],{"type":57,"tag":91,"props":132,"children":133},{},[134],{"type":63,"value":135},"Tool use problems",{"type":63,"value":137}," — wrong tool selected, malformed invocation, or poor handling of the tool response",{"type":57,"tag":87,"props":139,"children":140},{},[141,146],{"type":57,"tag":91,"props":142,"children":143},{},[144],{"type":63,"value":145},"Trajectory problems",{"type":63,"value":147}," — the agent took an inefficient path, got stuck in a loop, or failed to complete its task",{"type":57,"tag":87,"props":149,"children":150},{},[151,156],{"type":57,"tag":91,"props":152,"children":153},{},[154],{"type":63,"value":155},"Instrumentation gaps",{"type":63,"value":157}," — the application is poorly traced, leaving visibility gaps that make it difficult to understand behavior",{"type":57,"tag":58,"props":159,"children":161},{"id":160},"steps",[162],{"type":63,"value":163},"Steps",{"type":57,"tag":66,"props":165,"children":166},{},[167],{"type":63,"value":168},"Default query budget: 1–3 orientation queries, 3–5 aggregate or sampling queries, 3–7 targeted drilldowns. Avoid more than 20 GraphQL calls without summarizing progress. Avoid reading more than 5–10 full span inputs or outputs. Prefer fewer, richer queries — use GraphQL aliases to batch independent lookups. Use available GraphQL recipes before writing new queries.",{"type":57,"tag":170,"props":171,"children":172},"ol",{},[173,200,210,220,230],{"type":57,"tag":87,"props":174,"children":175},{},[176,181,183,190,192,198],{"type":57,"tag":91,"props":177,"children":178},{},[179],{"type":63,"value":180},"Orient",{"type":63,"value":182}," — Read ",{"type":57,"tag":184,"props":185,"children":187},"code",{"className":186},[],[188],{"type":63,"value":189},"\u002Fphoenix\u002Fagent-start.md",{"type":63,"value":191},". Use the ",{"type":57,"tag":184,"props":193,"children":195},{"className":194},[],[196],{"type":63,"value":197},"phoenix-gql",{"type":63,"value":199}," CLI to get a compact project overview in one aliased query: trace count, span count, latency quantiles, token totals, annotation names, slow traces, high-token traces, error spans.",{"type":57,"tag":87,"props":201,"children":202},{},[203,208],{"type":57,"tag":91,"props":204,"children":205},{},[206],{"type":63,"value":207},"Select",{"type":63,"value":209}," — Choose a representative sample. Prioritize slow traces, high-token traces, and errored traces; include a few normal traces for comparison. Prefer diversity over volume.",{"type":57,"tag":87,"props":211,"children":212},{},[213,218],{"type":57,"tag":91,"props":214,"children":215},{},[216],{"type":63,"value":217},"Open-code",{"type":63,"value":219}," — For each trace, write free-form notes on problems, surprises, and incorrect behaviors. Focus on the first failure in a trace, since upstream errors often cause downstream issues. Note independent downstream failures only when they reveal a separate root cause.",{"type":57,"tag":87,"props":221,"children":222},{},[223,228],{"type":57,"tag":91,"props":224,"children":225},{},[226],{"type":63,"value":227},"Axial-code",{"type":63,"value":229}," — Cluster your notes into named failure categories. Let categories emerge from the data. Distinguish exact counts from sampled or estimated counts.",{"type":57,"tag":87,"props":231,"children":232},{},[233,238],{"type":57,"tag":91,"props":234,"children":235},{},[236],{"type":63,"value":237},"Summarize",{"type":63,"value":239}," — Report findings using the output format below.",{"type":57,"tag":58,"props":241,"children":243},{"id":242},"observation-journal",[244],{"type":63,"value":245},"Observation Journal",{"type":57,"tag":66,"props":247,"children":248},{},[249],{"type":63,"value":250},"Output this table inline in the conversation as you work — it stays compact by design and needs to be visible for axial-coding. Only write it to a file if the user asks for exhaustive analysis across a large number of traces.",{"type":57,"tag":66,"props":252,"children":253},{},[254,256,262,264,270],{"type":63,"value":255},"For each inspected trace, add a row. Keep ",{"type":57,"tag":184,"props":257,"children":259},{"className":258},[],[260],{"type":63,"value":261},"observations",{"type":63,"value":263}," free-form. Fill in ",{"type":57,"tag":184,"props":265,"children":267},{"className":266},[],[268],{"type":63,"value":269},"tentative_category",{"type":63,"value":271}," only when the pattern is clear — leave it blank otherwise.",{"type":57,"tag":273,"props":274,"children":275},"table",{},[276],{"type":57,"tag":277,"props":278,"children":279},"thead",{},[280],{"type":57,"tag":281,"props":282,"children":283},"tr",{},[284,290,294],{"type":57,"tag":285,"props":286,"children":287},"th",{},[288],{"type":63,"value":289},"trace_id",{"type":57,"tag":285,"props":291,"children":292},{},[293],{"type":63,"value":261},{"type":57,"tag":285,"props":295,"children":296},{},[297],{"type":63,"value":269},{"type":57,"tag":66,"props":299,"children":300},{},[301],{"type":63,"value":302},"Use this journal as the input to axial-coding.",{"type":57,"tag":58,"props":304,"children":306},{"id":305},"output-format",[307],{"type":63,"value":308},"Output Format",{"type":57,"tag":83,"props":310,"children":311},{},[312,322,332],{"type":57,"tag":87,"props":313,"children":314},{},[315,320],{"type":57,"tag":91,"props":316,"children":317},{},[318],{"type":63,"value":319},"Analysis scope",{"type":63,"value":321}," — brief summary of what was analyzed: number of traces examined, time range if relevant, any filters applied",{"type":57,"tag":87,"props":323,"children":324},{},[325,330],{"type":57,"tag":91,"props":326,"children":327},{},[328],{"type":63,"value":329},"Findings table",{"type":63,"value":331}," — one row per issue category with: label, short description, occurrence count, one or two representative span (or trace) links — see Linking to Findings below",{"type":57,"tag":87,"props":333,"children":334},{},[335,340],{"type":57,"tag":91,"props":336,"children":337},{},[338],{"type":63,"value":339},"Recommendations",{"type":63,"value":341}," — for each issue, a concrete suggested fix (prompt change, parameter adjustment, tool fix, instrumentation improvement, etc.), if one can be identified",{"type":57,"tag":58,"props":343,"children":345},{"id":344},"linking-to-findings",[346],{"type":63,"value":347},"Linking to Findings",{"type":57,"tag":66,"props":349,"children":350},{},[351],{"type":63,"value":352},"Prefer linking to the specific span that exhibits the issue over the parent trace whenever possible — span links land the user on the exact node, while a trace link forces them to hunt for the relevant span. Fall back to a trace link only when no single span captures the issue (e.g., a trajectory problem spanning many spans).",{"type":57,"tag":66,"props":354,"children":355},{},[356,358,364,366,372,374,379,381,387,389,395,397,403,405,411],{"type":63,"value":357},"Use Phoenix's root-relative redirect URLs with the OpenTelemetry IDs returned by GraphQL — no project lookup required. Read the hex OTel IDs from ",{"type":57,"tag":184,"props":359,"children":361},{"className":360},[],[362],{"type":63,"value":363},"Span.spanId",{"type":63,"value":365}," and ",{"type":57,"tag":184,"props":367,"children":369},{"className":368},[],[370],{"type":63,"value":371},"Trace.traceId",{"type":63,"value":373},", ",{"type":57,"tag":91,"props":375,"children":376},{},[377],{"type":63,"value":378},"not",{"type":63,"value":380}," the ",{"type":57,"tag":184,"props":382,"children":384},{"className":383},[],[385],{"type":63,"value":386},"id",{"type":63,"value":388}," field (which is a Relay node ID and will not resolve). A ",{"type":57,"tag":184,"props":390,"children":392},{"className":391},[],[393],{"type":63,"value":394},"Span",{"type":63,"value":396}," has no ",{"type":57,"tag":184,"props":398,"children":400},{"className":399},[],[401],{"type":63,"value":402},"traceId",{"type":63,"value":404}," field — read it via the nested ",{"type":57,"tag":184,"props":406,"children":408},{"className":407},[],[409],{"type":63,"value":410},"trace { traceId }",{"type":63,"value":412},":",{"type":57,"tag":83,"props":414,"children":415},{},[416,427],{"type":57,"tag":87,"props":417,"children":418},{},[419,421],{"type":63,"value":420},"Span: ",{"type":57,"tag":184,"props":422,"children":424},{"className":423},[],[425],{"type":63,"value":426},"[short description](\u002Fredirects\u002Fspans\u002F\u003CspanId>)",{"type":57,"tag":87,"props":428,"children":429},{},[430,432],{"type":63,"value":431},"Trace: ",{"type":57,"tag":184,"props":433,"children":435},{"className":434},[],[436],{"type":63,"value":437},"[short description](\u002Fredirects\u002Ftraces\u002F\u003CtraceId>)",{"type":57,"tag":58,"props":439,"children":441},{"id":440},"caveats-and-pitfalls",[442],{"type":63,"value":443},"Caveats and Pitfalls",{"type":57,"tag":83,"props":445,"children":446},{},[447,452,457],{"type":57,"tag":87,"props":448,"children":449},{},[450],{"type":63,"value":451},"Existing evals and annotations are useful signal, but treat them as one input among many — they may be incomplete or incorrect.",{"type":57,"tag":87,"props":453,"children":454},{},[455],{"type":63,"value":456},"Span status codes are not a reliable proxy for whether an error actually occurred. An exception in the output may be expected behavior; a success status code may mask an error message buried in the span attributes.",{"type":57,"tag":87,"props":458,"children":459},{},[460],{"type":63,"value":461},"Do not overload your context by reading too many full spans. Read enough to support meaningful recommendations, then stop.",{"items":463,"total":543},[464,478,490,496,508,519,533],{"slug":465,"name":465,"fn":466,"description":467,"org":468,"tags":469,"stars":23,"repoUrl":24,"updatedAt":477},"annotate-spans","annotate LLM spans and traces","Write effective, consistent annotations on LLM\u002Fagent spans and traces, and coach the user on annotation practice. Load this whenever you are about to record structured feedback with the `batch_span_annotate` tool, or when the user asks how to annotate, label, score, or review spans\u002Ftraces, build a failure taxonomy, or set up human\u002FLLM review. Do NOT load for: pure analysis with no intent to save feedback (use debug-trace), latency or cost statistics, or prompt authoring (use playground).\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[470,472,475,476],{"name":471,"slug":35,"type":16},"Evals",{"name":473,"slug":474,"type":16},"LLM","llm",{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},"2026-07-12T08:08:14.140984",{"slug":34,"name":34,"fn":479,"description":480,"org":481,"tags":482,"stars":23,"repoUrl":24,"updatedAt":489},"reason about Phoenix dataset structure","Understand what a Phoenix dataset is and reason well about its examples, outputs, splits, and how it feeds evaluators and experiments. Load this whenever a dataset is in view or the user asks what a dataset is, how splits work, what an output \"means\", or how datasets relate to experiments and evals. This skill governs the judgment; any tool descriptions govern the mechanics.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[483,486,488],{"name":484,"slug":485,"type":16},"Data Analysis","data-analysis",{"name":487,"slug":34,"type":16},"Datasets",{"name":471,"slug":35,"type":16},"2026-07-12T08:08:21.695457",{"slug":4,"name":4,"fn":5,"description":6,"org":491,"tags":492,"stars":23,"repoUrl":24,"updatedAt":25},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[493,494,495],{"name":21,"slug":22,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},{"slug":497,"name":497,"fn":498,"description":499,"org":500,"tags":501,"stars":23,"repoUrl":24,"updatedAt":507},"evaluators","author and refine Phoenix evaluators","Author or refine a Phoenix evaluator — code or LLM-as-a-judge — that scores a run's output. Trigger when the user wants to create a new evaluator, improve an existing one's logic or rubric, choose labels, or decide what to measure on a dataset or experiment. Do NOT trigger on: (1) manual prompt drafting (use `playground`), (2) running or comparing experiments themselves (use `experiments`), (3) cross-trace failure diagnosis with no evaluator in scope (use `debug-trace`).\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[502,503,504],{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":505,"slug":506,"type":16},"Testing","testing","2026-07-31T05:58:09.13624",{"slug":509,"name":509,"fn":510,"description":511,"org":512,"tags":513,"stars":23,"repoUrl":24,"updatedAt":518},"experiments","run and compare dataset-backed experiments","Run, read, and compare dataset-backed experiments to find evidence that a prompt or pipeline is improving. Trigger when the user wants to iterate over a dataset with experiments, compare experiment runs, read experiment quality\u002Flatency\u002Fcost, or decide whether a change actually helped. Running a prompt over a dataset is implicitly an experiment — load this skill when dataset-backed work begins, before authoring evaluators for the experiment and before starting the recorded run, not only when reading results. Do NOT trigger on: (1) manual prompt drafting with no dataset-backed evaluation in scope (use `playground`), (2) authoring or refining an evaluator's logic or rubric (use `evaluators`), (3) cross-trace failure diagnosis with no experiment in scope (use `debug-trace`).\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[514,515,516,517],{"name":487,"slug":34,"type":16},{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":505,"slug":506,"type":16},"2026-07-12T08:08:11.691477",{"slug":520,"name":520,"fn":521,"description":522,"org":523,"tags":524,"stars":23,"repoUrl":24,"updatedAt":532},"phoenix-graphql","query Phoenix API with GraphQL","Write efficient GraphQL queries against the Phoenix API. Load this skill in two cases: (1) before composing any non-trivial GraphQL query yourself for data analysis (via the `phoenix-gql` bash command) — it contains schema entrypoints and patterns that eliminate the need for introspection; (2) when the user asks for help writing GraphQL queries for their own scripts, tools, or integrations against Phoenix — it covers the endpoint, authentication, and client examples.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[525,528,529],{"name":526,"slug":527,"type":16},"Analytics","analytics",{"name":484,"slug":485,"type":16},{"name":530,"slug":531,"type":16},"GraphQL","graphql","2026-07-12T08:08:17.163493",{"slug":534,"name":534,"fn":535,"description":536,"org":537,"tags":538,"stars":23,"repoUrl":24,"updatedAt":542},"playground","author and iterate on prompts in Phoenix","Author, edit, or iterate on prompts in the Phoenix prompt playground, including running experiments over a dataset. Load before any playground tool call, including single-shot prompt rewrites.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[539,540,541],{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":505,"slug":506,"type":16},"2026-07-12T08:08:12.920792",8,{"items":545,"total":669},[546,553,559,565,571,578,584,590,601,619,637,647],{"slug":465,"name":465,"fn":466,"description":467,"org":547,"tags":548,"stars":23,"repoUrl":24,"updatedAt":477},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[549,550,551,552],{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},{"slug":34,"name":34,"fn":479,"description":480,"org":554,"tags":555,"stars":23,"repoUrl":24,"updatedAt":489},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[556,557,558],{"name":484,"slug":485,"type":16},{"name":487,"slug":34,"type":16},{"name":471,"slug":35,"type":16},{"slug":4,"name":4,"fn":5,"description":6,"org":560,"tags":561,"stars":23,"repoUrl":24,"updatedAt":25},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[562,563,564],{"name":21,"slug":22,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},{"slug":497,"name":497,"fn":498,"description":499,"org":566,"tags":567,"stars":23,"repoUrl":24,"updatedAt":507},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[568,569,570],{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":505,"slug":506,"type":16},{"slug":509,"name":509,"fn":510,"description":511,"org":572,"tags":573,"stars":23,"repoUrl":24,"updatedAt":518},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[574,575,576,577],{"name":487,"slug":34,"type":16},{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":505,"slug":506,"type":16},{"slug":520,"name":520,"fn":521,"description":522,"org":579,"tags":580,"stars":23,"repoUrl":24,"updatedAt":532},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[581,582,583],{"name":526,"slug":527,"type":16},{"name":484,"slug":485,"type":16},{"name":530,"slug":531,"type":16},{"slug":534,"name":534,"fn":535,"description":536,"org":585,"tags":586,"stars":23,"repoUrl":24,"updatedAt":542},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[587,588,589],{"name":471,"slug":35,"type":16},{"name":473,"slug":474,"type":16},{"name":505,"slug":506,"type":16},{"slug":591,"name":591,"fn":592,"description":593,"org":594,"tags":595,"stars":23,"repoUrl":24,"updatedAt":600},"span-coding","analyze and code Phoenix spans","Open-code Phoenix spans with PXI-owned notes, recover those notes for axial coding, and promote stable categories into structured annotations. Load this when analyzing spans to discover failure patterns before a taxonomy exists.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[596,597,598,599],{"name":21,"slug":22,"type":16},{"name":473,"slug":474,"type":16},{"name":18,"slug":19,"type":16},{"name":14,"slug":15,"type":16},"2026-07-12T08:08:19.597239",{"slug":602,"name":602,"fn":603,"description":604,"org":605,"tags":606,"stars":616,"repoUrl":617,"updatedAt":618},"arize-admin","manage Arize enterprise user access","Manages Arize users, organizations, spaces, projects, roles, role bindings, resource restrictions, and API keys via the ax CLI. Use for enterprise admin workflows: inviting and offboarding users, onboarding new teams, creating custom roles for SAML\u002FSSO mappings, assigning roles to users, restricting project-level access, and managing service keys for multi-tenant architectures. Covers ax users, ax organizations, ax spaces, ax projects, ax roles, ax role-bindings, and ax api-keys.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[607,610,613],{"name":608,"slug":609,"type":16},"CLI","cli",{"name":611,"slug":612,"type":16},"Operations","operations",{"name":614,"slug":615,"type":16},"Permissions","permissions",38,"https:\u002F\u002Fgithub.com\u002FArize-ai\u002Farize-skills","2026-07-22T05:37:21.991338",{"slug":620,"name":620,"fn":621,"description":622,"org":623,"tags":624,"stars":616,"repoUrl":617,"updatedAt":636},"arize-ai-provider-integration","manage Arize AI provider integrations","Creates, reads, updates, and deletes Arize AI integrations that store LLM provider credentials used by evaluators and other Arize features. Supports any LLM provider (e.g. OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Vertex AI, Gemini, NVIDIA NIM). Use when the user mentions AI integration, LLM provider credentials, create integration, list integrations, update credentials, delete integration, or connecting an LLM provider to Arize.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[625,627,630,633,634],{"name":626,"slug":33,"type":16},"Anthropic",{"name":628,"slug":629,"type":16},"Azure","azure",{"name":631,"slug":632,"type":16},"Integrations","integrations",{"name":473,"slug":474,"type":16},{"name":635,"slug":42,"type":16},"OpenAI","2026-07-22T05:37:23.90468",{"slug":638,"name":638,"fn":639,"description":640,"org":641,"tags":642,"stars":616,"repoUrl":617,"updatedAt":646},"arize-annotation","manage Arize annotation workflows","Creates and manages annotation configs (categorical, continuous, freeform label schemas) and annotation queues (human review workflows) on Arize. Applies human annotations to project spans via the Python SDK. Use when the user mentions annotation config, annotation queue, label schema, human feedback, bulk annotate spans, update_annotations, labeling queue, annotate record, or human review.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[643,644,645],{"name":484,"slug":485,"type":16},{"name":473,"slug":474,"type":16},{"name":18,"slug":19,"type":16},"2026-07-22T05:37:19.010776",{"slug":648,"name":648,"fn":649,"description":650,"org":651,"tags":652,"stars":616,"repoUrl":617,"updatedAt":668},"arize-compliance-audit","audit AI agents for regulatory compliance","INVOKE THIS SKILL when auditing an AI agent or LLM app for regulatory compliance. Covers EU AI Act, GPAI Code of Practice, GDPR, NIST AI RMF, Colorado AI Act, HIPAA, and ISO 42001. Scans the codebase for compliance gaps, cross-references Arize instrumentation for audit trail coverage, and produces an actionable remediation checklist tailored to the selected frameworks.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[653,656,659,662,665],{"name":654,"slug":655,"type":16},"Audit","audit",{"name":657,"slug":658,"type":16},"Compliance","compliance",{"name":660,"slug":661,"type":16},"GDPR","gdpr",{"name":663,"slug":664,"type":16},"Legal","legal",{"name":666,"slug":667,"type":16},"Security","security","2026-07-19T05:39:42.632738",23]