🪢 Open source agent evals & observability: Trace, evaluate, and improve LLM applications with one open platform.
TypeScript analyticsautogenevaluationlangchainlarge-language-modelsllama-indexllmllm-evaluationllm-observabilityllmopsmonitoringobservabilityopen-sourceopenaiplaygroundprompt-engineeringprompt-managementself-hostedycombinator
chore(assistant): refresh embedded Langfuse skills (#17766)
langfuse-bote995dc730ff3
diff --git a/packages/langfuse-skills/src/generated/skills.js b/packages/langfuse-skills/src/generated/skills.jsindex fecfcb23b3..f080237a5d 100644--- a/packages/langfuse-skills/src/generated/skills.js+++ b/packages/langfuse-skills/src/generated/skills.js@@ -58,13 +58,6 @@ exports.LANGFUSE_SKILLS = [instructions:'# Langfuse Prompt Migration\n\nMigrate hardcoded prompts into Langfuse-managed prompts. The API mechanics (`create_prompt`, `get_prompt`, `.compile()`, linking to traces) are in the docs — fetch them at execution time.\n\n## Prerequisites\n\nVerify credentials exist — check presence only, never print the secret key (its value would land in the agent\'s context and transcripts):\n\n```bash\n[ -n "$LANGFUSE_PUBLIC_KEY" ] && echo "public key: set" || echo "public key: missing"\n[ -n "$LANGFUSE_SECRET_KEY" ] && echo "secret key: set" || echo "secret key: missing"\n[ -n "${LANGFUSE_BASE_URL:-${LANGFUSE_HOST:-}}" ] && echo "base url: set" || echo "base url: missing"\n```\n\nUse `LANGFUSE_BASE_URL` for current SDKs. If only `LANGFUSE_HOST` is set, export `LANGFUSE_BASE_URL="$LANGFUSE_HOST"`. If a CLI expects `LANGFUSE_HOST` and only `LANGFUSE_BASE_URL` is set, export `LANGFUSE_HOST="$LANGFUSE_BASE_URL"`.\n\nIf credentials or the base URL are missing, ask the user to set them in their shell or a `.env` file. Do not ask them to paste secret keys into chat.\n\n## 1. Inventory every prompt (before writing any code)\n\nFor each prompt, record:\n\n- **Name**: lowercase, hyphenated (e.g. `chat-assistant`)\n- **Source file**: where the prompt text lives\n- **Code file to refactor**: the file that USES the prompt. For asset files (`.txt`/`.yaml`/`.md`), this is the file that loads the asset, not the asset itself\n- **Type**: `chat` (message array) or `text` (plain string)\n- **Variables**: values interpolated in, converted to `{{var}}`\n- **Content**: the actual text to upload\n\nBefore choosing `text` or `chat`, fetch and follow [Chat vs Text Prompts](https://langfuse.com/docs/prompt-management/data-model#text-vs-chat-prompts). Do not default unrelated prompt flows to one prompt type.\n\nPrompts typically live in OpenAI message arrays, Anthropic system arguments, LangChain prompt templates, Vercel AI system/prompt fields, and raw multi-line strings near LLM calls.\n\n## 2. Convert templating and decide structure\n\n**Variable syntax:** Langfuse substitutes only double-brace `{{var}}`. Convert every single-brace form during upload — `{var}`, `${var}`, f-string `{var}`, `.format(var=...)`, and string concatenation all become `{{var}}`. Uploading `{var}` will silently fail to substitute.\n\n**Complex templates:** Langfuse has no conditionals, loops, or filters. If the code uses them (e.g. Jinja `{% if %}`/`{% for %}`), either pre-compute the value in code and pass a plain `{{variable}}` (recommended), or store the raw template and compile client-side — which loses Playground preview and UI experiments. See https://langfuse.com/docs/prompt-management/features/variables and the [external templating FAQ](https://langfuse.com/faq/all/using-external-templating-libraries).\n\n**What to make a variable vs. keep hardcoded:**\n\n| Make variable | Keep hardcoded |\n|---------------|----------------|\n| User-specific (`{{user_name}}`) | Output format instructions |\n| Dynamic content (`{{context}}`) | Safety guardrails |\n| Per-request (`{{query}}`) | Persona / personality |\n| Environment-specific (`{{company_name}}`) | Static examples |\n\n**Naming:** lowercase-hyphenated, feature-based (`document-summarizer`), hierarchical for related prompts (`support/triage`), prefix subprompts with `_` (`_base-personality`).\n\nBefore extracting subprompts, fetch and follow [Prompt Composability](https://langfuse.com/docs/prompt-management/features/composability). Use composition for the reuse and shared-maintenance cases described there, not merely to decompose one coherent prompt flow.\n\n## 3. Present the plan, then create and refactor\n\nWhen the user explicitly requests a migration and credentials work, treat that request as authorization to create prompts and refactor the call sites. Present the inventory and plan as a progress update, then continue.\n\nAsk only when a materially different design choice would change behavior, required credentials are unavailable, or destructive cleanup needs approval. Do not stop merely to confirm names, prompt types, or optional tracing.\n\nThen:\n\n- Create the prompts (label migrated prompts `production` — they\'re already live) and refactor call sites to fetch each prompt from Langfuse and compile its variables in. The SDK calls differ across Python and JS/TS — fetch the current docs: https://langfuse.com/docs/prompt-management/get-started\n- Fetch by the `production` label\n- If the codebase already has Langfuse tracing (decorators, an instrumented client, or manual spans), link prompts so you can see which version produced each response. See https://langfuse.com/docs/prompt-management/features/link-to-traces\n\n## 4. Verify\n\n- All prompts created with the `production` label; code fetches with `label="production"`\n- Variables and subprompts compile without errors\n- Application behavior is unchanged\n- Generations show the linked prompt in the UI (if tracing enabled)',},- {- name: "langfuse-sdk-upgrade",- description:- "Upgrade Langfuse SDKs and application instrumentation while preserving trace attributes across observations. Use for Python or JS/TS SDK migrations, including the application side of a v4 platform migration.",- instructions:- "# Langfuse SDK upgrade\n\n## Sources of truth\n\nDetermine every installed Langfuse SDK major version, then fetch each applicable leaf guide in full before editing:\n\n- [SDK upgrade paths](https://langfuse.com/docs/observability/sdk/upgrade-path)\n- [Python v3 to v4](https://langfuse.com/docs/observability/sdk/upgrade-path/python-v3-to-v4)\n- [JS/TS v4 to v5](https://langfuse.com/docs/observability/sdk/upgrade-path/js-v4-to-v5)\n- [Instrumentation and attribute propagation](https://langfuse.com/docs/observability/sdk/instrumentation#add-attributes)\n- [Sessions and session-level metrics](https://langfuse.com/docs/observability/features/sessions)\n- [Direct OpenTelemetry ingestion](https://langfuse.com/integrations/native/opentelemetry)\n\nFollow every intermediate guide when the installed version is more than one major behind. Use the current docs for implementation details; do not copy examples from this file.\n\n## Workflow\n\n1. Inventory every SDK, integration package, direct OpenTelemetry exporter, initialization site, instrumentation wrapper, lockfile, worker, script, and test that can emit Langfuse data.\n2. Find every source of correlating attributes, including `session_id`/`sessionId`, `user_id`/`userId`, tags, metadata, version, environment, and trace name. Search for the values and surrounding application concepts, not only removed SDK method names.\n3. Apply every relevant item from the exact version-specific guides. Preserve each correlating attribute by establishing its documented propagation scope before any observation-producing call that must inherit it.\n\n\n## Completion report\n\nReport the versions before and after, changed instrumentation paths, attribute sources and propagation scopes, validation performed, inspected trace or session, and any remaining blocked verification.",- },{name: "langfuse-setting-up-evals",description:@@ -79,13 +72,6 @@ exports.LANGFUSE_SKILLS = [instructions:"# Skill Feedback\n\nUse this only for feedback about the skill itself, not Langfuse the product or a user's application data.\n\n1. Draft concise feedback using the following structure:\n - **Describe your idea or feedback** (required)\n A clear description of what went wrong or what could be improved. Include:\n - What the user was trying to do (as `goal`)\n - What the skill did vs what was expected\n - Any specific instructions that were incorrect or missing\n\n - **What would the ideal outcome look like?** (optional)\n What the correct behavior or guidance should be.\n\n If feedback targets existing skill, reference in `target`.\n2. If the user wants a reply, ask them to include an email address in `feedback`; use only an address they explicitly provide.\n3. Show every submitted field exactly as it will be sent and ask for explicit permission. Do not submit without approval.\n\n## Submission options\n\nUse the first available option, or the option the user prefers:\n\n1. **Authenticated Langfuse MCP, CLI, or Public API** — prefer the `submitFeedback` tool on the Langfuse MCP server. If it is unavailable, discover the current feedback operation with the Langfuse CLI schema/help and submit through the authenticated Public API. Do not ask users to paste credentials into chat.\n2. **Langfuse Docs MCP** — use its unauthenticated `submitFeedback` tool when no authenticated Langfuse interface is available.\n3. **GitHub issue or discussion** — if no MCP/CLI/API is available, or the user wants a public, trackable thread, provide a prefilled discussion link for the user to submit: `https://github.com/langfuse/skills/discussions/new?category=ideas-improvements&title=<url-encoded title>&body=<url-encoded body>`. Use `https://github.com/langfuse/skills/issues/new` if they prefer an issue.\n\nReport the Langfuse correlation ID or GitHub URL. If submission fails, give a safe error without exposing credentials or request bodies.",},- {- name: "langfuse-evaluator-upgrade",- description:- "Upgrade legacy trace-level or dataset-item LLM-as-a-Judge evaluators to observation-level or experiment evaluators. Use when a project needs the coding handoff for evaluator migration during the v4 transition.",- instructions:- "# Evaluator upgrade\n\nUse the [evaluator migration guide](https://langfuse.com/faq/all/llm-as-a-judge-migration),\n[v4 overview](https://langfuse.com/docs/v4), and current [Evaluation Rules](https://api.reference.langfuse.com/#tag/unstableevaluationrules)\nand [Evaluators](https://api.reference.langfuse.com/#tag/unstableevaluators) API schemas as\nthe sources of truth. Read the current unstable schema before using it: only the unstable\nEvaluation Rules API returns legacy `trace` and `dataset` targets; new rules use observation\nor experiment targets.\n\n## Inventory and scope\n\n- Page through all rules and fetch each referenced evaluator. Record target, status, filters, mappings/JSONPaths, sampling, delay, time scope, evaluator, and score name.\n- If a bulk page fails, retry with `limit=1` to isolate unreadable entries; report any page that still fails as a blocker and do not claim a complete inventory.\n- Separate active, inactive, and blocked rules. Migrate only active rules that depend on new/live data. Historical-only rules (`EXISTING` without `NEW`) may be ignored; deactivation or deletion is optional and requires approval.\n- Migrate only legacy trace and dataset targets here. Do not touch existing observation, experiment, or event rules. If the unstable API cannot mutate a legacy rule, return the exact UI action.\n- Show the complete inventory and a consolidated retain/delete decision before changing project configuration.\n\n## Trace to observation\n\nIn v4, a trace groups observations and has no separate trace input/output target. For each\nretained trace rule, select one stable observation and verify its own fields:\n\n1. Prefer an existing root or parent observation that already contains the required result.\n2. If the rule used overall trace input/output, confirm that the previous SDK upgrade has already put those values on the root observation.\n3. If it mixed trace context with one observation, propagate only the required context or filter attributes to that observation.\n4. If required values span multiple observations, add a dedicated evaluation observation only when no existing observation can own them. Do not serialize the whole trace into metadata.\n\nUse a stable observation name/type and root status where relevant. The selector should ideally match at\nmost one observation per trace when the legacy rule produced one score per trace. If it matches multiple observation, convey to the user that the rule will produce one score per observation. Keep names\nand types stable across executions.\n\nPreserve the evaluator-template: prompt, model, output, score name, sampling, delay. Time scope, filters, mappings, and JSONPaths need to be compatible with the new format. A legacy trace\nname filter becomes `traceName`; it does not replace the observation selector. Every variable\nand filter must exist on the selected observation, with trace attributes propagated according\nto the applicable SDK or OpenTelemetry guide.\n\nInspect the resolved SDK/OTel path and a representative new trace; do not infer readiness\nfrom a dependency declaration alone. Complete application instrumentation first when needed.\nKeep deprecated trace-I/O calls only while an active legacy rule requires them, and remove\nthem after the successor is verified and the legacy rule is disabled.\n\n## Dataset item to experiment\n\nRead the legacy `dataset` rule and referenced evaluator, then create an `experiment` rule with\nthe same evaluator, score name, sampling, and mappings. Use the experiment target's current\ndataset filter and dataset ID from the unstable schema.\nAs of the current API, this is `datasetId` with IDs from the v2 dataset endpoint. Translate\ndataset-item expected output and metadata to the corresponding experiment-item fields, and\npreserve input mappings and JSONPaths against the experiment-item shape.\n\nFor dataset runs, the root observation carries the run's input and output, so mapping legacy\ntrace input/output to observation input/output is safe and needs no evaluator-specific\ninstrumentation change.\n\nKeep the legacy rule active while the project still uses low-level dataset-run APIs. Disable\nit after the project emits experiment context; leaving both active can create duplicate\nscores. Review any mapping that relies on a named observation or assumes a one-to-one input.\n\n## Cut over and report\n\n- Create the successor disabled when supported, validate it on new data, then enable it with approval. A brief overlap can create duplicate scores.\n- Disable the legacy rule only after the successor is verified. Do not backfill historical data or delete historical scores.\n- Verify the selected observation exists once when required, contains every mapped value and filter, and excludes sibling observations. Do not claim live verification without project access or a runnable path.\n\nReturn a concise per-rule table with the legacy rule, target, selector, mappings/filters,\nrequired code changes, validation result, and remaining project actions or blockers.",- },{name: "langfuse-user-feedback",description:@@ -98,6 +84,6 @@ exports.LANGFUSE_SKILLS = [description:"Prepare an application and Langfuse project for v4 by upgrading ingestion and API usage, migrating legacy evaluators, and moving exports to the enriched observation schema.",instructions:- "# Langfuse v4 project migration\n\n## Sources of truth\n\nFetch only the pages needed for the surfaces found:\n\n- [V4 overview](https://langfuse.com/docs/v4) and [compatibility](https://langfuse.com/docs/compatibility)\n- [SDK upgrade paths](https://langfuse.com/docs/observability/sdk/upgrade-path)\n- [Custom ingestion migration](https://langfuse.com/integrations/native/opentelemetry/migration-to-v4)\n- [Deprecated API migration](https://langfuse.com/faq/all/deprecated-api-migration)\n- [Evaluator migration](https://langfuse.com/faq/all/llm-as-a-judge-migration)\n- [Evaluation Rules](https://api.reference.langfuse.com/#tag/unstableevaluationrules) and [Evaluators](https://api.reference.langfuse.com/#tag/unstableevaluators) APIs\n- Export migrations: [Blob Storage](https://langfuse.com/docs/api-and-data-platform/features/export-to-blob-storage#upgrade-path), [Mixpanel](https://langfuse.com/integrations/analytics/mixpanel#migrate-export-source), and [PostHog](https://langfuse.com/integrations/analytics/posthog#migrate-export-source)\n\nDiscover unstable schemas before use.\n\n## Choose the execution mode\n\n- Confirm the target host and project before project reads or writes. Never request secrets in chat or commit them.\n- Prefer an available project interface. Otherwise use the [CLI](https://langfuse.com/docs/api-and-data-platform/features/cli), starting with `api __schema` and action `--help`.\n- Without project access, enter **code-only mode**: complete verifiable repository work, do not infer evaluator or export state, and mark every project-dependent result blocked.\n- Without codebase access, inspect the project and return an exact code handoff; do not mark repository work ready.\n\n## Migrate the codebase\n\n- Inventory every Langfuse SDK, integration, OTEL exporter, initialization site, lockfile, raw request, generated client, script, notebook, and CI call.\n- Upgrade to the latest stable SDK major required by the current docs and apply every applicable breaking change. Record both declared and resolved versions; update an existing lockfile.\n- Find every source of correlating attributes, including session and user IDs, tags, metadata, version, environment, and trace name; do not search only for removed SDK methods.\n- Put overall input/output on the root observation. Establish the documented propagation scope before observation-producing calls so every applicable child receives the attributes needed for filtering and aggregation, including the session ID on cost-bearing generations.\n- For raw `/api/public/ingestion`, use the current Langfuse SDK in Python or JS/TS. For other languages, use the language's native OpenTelemetry API and follow the custom-ingestion guide.\n- When replacing synchronous ingestion, assess buffering, retries, flushing, shutdown, and error propagation. Do not claim identical delivery semantics without verification.\n- For other deprecated APIs, migrate the path, parameters, pagination, filters, field groups, response parsing, and downstream consumer together using the deprecated-API guide.\n- Check self-hosted compatibility before replacing calls; report code targeting a v3 server as blocked on the server upgrade.\n\n## Migrate evaluators\n\n- Inventory all active evaluators and rules, inspect each referenced definition, and report inactive rules without reactivating them.\n- Check the Evaluators UI on the confirmed target host for active **Legacy** rows ([open in Langfuse Cloud](https://cloud.langfuse.com/project/~/evals)); public interfaces may not expose every legacy target.\n- For each inspected active legacy rule, record its filters, sampling, mappings, representative payloads, one observation or experiment successor, required code changes, and cutover test.\n- Label this a **project-verified contract** only when the rule and representative observations were inspected. Label code-only suggestions **candidate targets**, never project state.\n- Consolidate all required evaluator variables and filter attributes onto the one target observation. Observation evaluators cannot read siblings or children.\n- Create or update the successor disabled, validate it on newly ingested data, obtain approval before enabling it, compare scores and logs, then disable rather than delete the legacy rule.\n\n### Deprecated trace input/output escape hatch\n\n- Default to removing deprecated trace input/output and migrating the evaluator to root-observation input/output.\n- Never add or retain `set_current_trace_io()`, `span.set_trace_io()`, `setActiveTraceIO()`, `span.setTraceIO()`, or equivalent compatibility merely because a legacy rule exists or project access is unavailable.\n- Offer compatibility only when automatic migration is currently impossible or the user explicitly insists on preserving the legacy evaluator unchanged after being warned that trace input/output is deprecated.\n- Even then, require explicit confirmation before changing code, mark the evaluator migration `manual action` or `blocked`, and document the condition for removing the escape hatch. Do not present it as the recommended migration.\n\n## Migrate exports\n\n- Inventory Blob Storage, Mixpanel, PostHog, and other sources in **Project Settings > Integrations**.\n- Follow each documented dual-source transition. Enabling dual output is additive; enriched-only stops legacy output and requires explicit downstream-owner confirmation.\n- Validate warehouse schemas, paths, joins, dashboards, transformations, alerts, and historical-data handling. Keep every consumer on one shape during dual mode.\n- Change only the source; preserve credentials, schedules, prefixes, formats, field groups, and secrets. Report the area `manual action` until downstream consumers are confirmed.\n\n## Validate and report\n\n- Test applicable hierarchy, root input/output, propagated attributes, public/release/environment behavior, API pagination and parsing, delivery semantics, and absence of deprecated calls.\n- Before production cutover, send representative traces from the migrated instrumentation to a non-production Langfuse project and inspect the resulting observations there. Mocked tests do not verify backend ingestion or project behavior.\n- On a session path, confirm the root and every applicable child observation carry the intended session ID and that session cost includes the cost-bearing children.\n- Re-read rules and integrations after writes. Preserve disabled legacy rules for rollback; never claim completion without checking the Evaluators UI on the target host for legacy rows.\n- The readiness report must contain exactly these seven rows, each marked `ready`, `changed`, `manual action`, or `blocked`: project access; SDK/instrumentation; trace evaluators; dataset evaluators; direct APIs; exports; verification/rollback.\n- In evaluator rows, separate contracts verified against project data from targets suggested only by code inspection. For every row not marked `ready`, include the blocker, next action, and a direct UI link when applicable.",+ "# Langfuse v4 project migration\n\n## Sources of truth\n\nFetch only the pages needed for the surfaces found:\n\n- [V4 upgrade guide](https://langfuse.com/faq/all/upgrade-to-langfuse-v4) and [compatibility](https://langfuse.com/docs/compatibility)\n- [SDK upgrade paths](https://langfuse.com/docs/observability/sdk/upgrade-path)\n- [Custom ingestion migration](https://langfuse.com/integrations/native/opentelemetry/migration-to-v4)\n- [Deprecated API migration](https://langfuse.com/faq/all/deprecated-api-migration)\n- [Evaluator migration](https://langfuse.com/faq/all/llm-as-a-judge-migration)\n- [Evaluation Rules](https://api.reference.langfuse.com/#tag/evaluationrules) and [Evaluators](https://api.reference.langfuse.com/#tag/unstableevaluators) APIs\n- Export migrations: [Blob Storage](https://langfuse.com/docs/api-and-data-platform/features/export-to-blob-storage#upgrade-path), [Mixpanel](https://langfuse.com/integrations/analytics/mixpanel#migrate-export-source), and [PostHog](https://langfuse.com/integrations/analytics/posthog#migrate-export-source)\n\nDiscover the current schemas before use.\n\n## Choose the execution mode\n\n- Confirm the target host and project before project reads or writes. Never request secrets in chat or commit them.\n- Prefer an available project interface. Otherwise use the [CLI](https://langfuse.com/docs/api-and-data-platform/features/cli), starting with `api __schema` and action `--help`.\n- Without project access, enter **code-only mode**: complete verifiable repository work, do not infer evaluator or export state, and mark every project-dependent result blocked.\n- Without codebase access, inspect the project and return an exact code handoff; do not mark repository work ready.\n\n## Migrate the codebase\n\n- Inventory every Langfuse SDK, integration, OTEL exporter, initialization site, lockfile, raw request, generated client, script, notebook, and CI call.\n- Upgrade to the latest stable SDK major required by the current docs and apply every applicable breaking change. Record both declared and resolved versions; update an existing lockfile.\n- Find every source of correlating attributes, including session and user IDs, tags, metadata, version, environment, and trace name; do not search only for removed SDK methods.\n- Put overall input/output on the root observation. Establish the documented propagation scope before observation-producing calls so every applicable child receives the attributes needed for filtering and aggregation, including the session ID on cost-bearing generations.\n- For raw `/api/public/ingestion`, use the current Langfuse SDK in Python or JS/TS. For other languages, use the language's native OpenTelemetry API and follow the custom-ingestion guide.\n- When replacing synchronous ingestion, assess buffering, retries, flushing, shutdown, and error propagation. Do not claim identical delivery semantics without verification.\n- For other deprecated APIs, migrate the path, parameters, pagination, filters, field groups, response parsing, and downstream consumer together using the deprecated-API guide.\n- Check self-hosted compatibility before replacing calls; report code targeting a v3 server as blocked on the server upgrade.\n\n## Migrate evaluators\n\n- Inventory all active evaluators and rules, inspect each referenced definition, and report inactive rules without reactivating them.\n- Check the Evaluators UI on the confirmed target host for active **Legacy** rows ([open in Langfuse Cloud](https://cloud.langfuse.com/project/~/evals)); public interfaces may not expose every legacy target.\n- For each inspected active legacy rule, record its filters, sampling, mappings, representative payloads, one observation or experiment successor, required code changes, and cutover test.\n- Label this a **project-verified contract** only when the rule and representative observations were inspected. Label code-only suggestions **candidate targets**, never project state.\n- Consolidate all required evaluator variables and filter attributes onto the one target observation. Observation evaluators cannot read siblings or children.\n- Create or update the successor disabled, validate it on newly ingested data, obtain approval before enabling it, compare scores and logs, then disable rather than delete the legacy rule.\n\n## Migrate exports\n\n- Inventory Blob Storage, Mixpanel, PostHog, and other sources in **Project Settings > Integrations**.\n- Follow each documented dual-source transition. Enabling dual output is additive; enriched-only stops legacy output and requires explicit downstream-owner confirmation.\n- Validate warehouse schemas, paths, joins, dashboards, transformations, alerts, and historical-data handling. Keep every consumer on one shape during dual mode.\n- Change only the source; preserve credentials, schedules, prefixes, formats, field groups, and secrets. Report the area `manual action` until downstream consumers are confirmed.\n\n## Validate and report\n\n- Test applicable hierarchy, root input/output, propagated attributes, public/release/environment behavior, API pagination and parsing, delivery semantics, and absence of deprecated calls.\n- Before production cutover, send representative traces from the migrated instrumentation to a non-production Langfuse project and inspect the resulting observations there. Mocked tests do not verify backend ingestion or project behavior.\n- On a session path, confirm the root and every applicable child observation carry the intended session ID and that session cost includes the cost-bearing children.\n- Re-read rules and integrations after writes. Preserve disabled legacy rules for rollback; never claim completion without checking the Evaluators UI on the target host for legacy rows.\n- The readiness report must contain exactly these seven rows, each marked `ready`, `changed`, `manual action`, or `blocked`: project access; SDK/instrumentation; trace evaluators; dataset evaluators; direct APIs; exports; verification/rollback.\n- In evaluator rows, separate contracts verified against project data from targets suggested only by code inspection. For every row not marked `ready`, include the blocker, next action, and a direct UI link when applicable.",},];