cekura-skills

Use when the user asks to create, update, or review a source-controlled Cekura JSON CI/CD test suite for a voice AI repository; create Tests-as-Code specs; turn a voice-agent code change into regression coverage; add deterministic Cekura voice tests to CI; or test an STT, LLM, TTS, VAD, interruption, idle-timer, DTMF, or call-lifecycle pipeline; or set up a CI gate that blocks a merge when the voice pipeline regresses. Inspects the repository before authoring a compact JSON suite, validates it safely with Cekura dry-run, and wires the workflow that runs it.

Turn a set of flagged production call logs into evaluator scenarios for a Cekura agent — cluster the failures into reproducible modes and create one scenario per cluster. Use when the user says "create scenarios from failed calls", "build evaluators for prod failures", "turn call logs into scenarios", "simulate the failures we saw in prod", "replay these bad calls as tests", "regression-test the agent on prod issues", or hands over a set of flagged call IDs to harden the agent against. The flagged set normally comes from `cekura-flag-call-log-failures`; if the user hasn't triaged yet, run that skill first. This skill takes the flagged calls as given — it does NOT re-mine or re-triage — clusters by failure mode, drafts one scenario per cluster, and optionally creates them via the scenarios API (or the call-logs → scenarios endpoint). Also supports a single-call fast path: given one call log ID (or an observe URL), reproduce just that call as one evaluator scenario.

Triage the last N production call logs for a Cekura agent against a set of KPIs / issues / goals, report the call IDs that hit each failure WITH the percentage of overall call logs affected, and distribute the remaining calls into a mutually-exclusive outcome taxonomy (e.g. not-answered / vetted / non-vetted caller-side / non-vetted agent-issue) with per-bucket percentages. Use when the user says "flag call log failures", "analyze the last N calls for issues", "what % of calls have <problem>", "which calls broke and how often", "find failing calls", "give me the breakdown of call outcomes", "what % of answered calls can be improved", or pastes an agent / project ID and lists the problems or KPIs to measure. Applies attribution rules so caller-side endings and recovered calls are NOT counted as agent failures. This is the upstream triage step that feeds `cekura-generate-scenarios` — it does NOT create scenarios.

Use when the user asks to "create a metric", "write a metric", "design a metric", "build a metric for", "evaluate agent performance", "measure call quality", "track a KPI", "add a workflow metric", "improve my metric", "fix a metric", "debug metric results", "set up quality scoring", or "what metrics do I need". Also relevant when discussing LLM judge prompts, custom code metrics, evaluation triggers, VALID_SKIP patterns, section extraction, or metric best practices for Cekura voice AI agents. Covers both creating new metrics and reviewing, iterating on, or troubleshooting existing ones.

Use when the user asks "what predefined metrics are available", "which built-in metrics should I use", "what does CSAT measure", "how does hallucination detection work", "what's the difference between Interruption Score and AI Interrupting User", "which metrics are free", "which metrics need audio", "configure silence threshold", "set up sentiment metric", or any question about Cekura's out-of-the-box metrics. Covers the full catalog of predefined metrics — what each does, costs, constraints, configuration options, and when to use each one.

Use to close the loop on agent quality — turn a failure signal into a verified fix. Triggers: "improve my agent", "self-improving agent", "auto-tune / iterate on my prompt", "fix my agent from test results" — and production-call bug fixing: "fix this prod call issue", "debug and fix call ID", "reproduce this production bug", "regression test before raising PR". Works on any stack: provider-dashboard agents (VAPI, Retell, ElevenLabs, Bland) AND agents whose config lives in the customer's own repo, database, or prompt registry with the provider agent created at deploy time, or with custom mock servers. Fixed safety invariants (must-fail-first reproduction, attestation, no-prod-in-loop) + a per-project capability manifest declaring where config lives and how to read/render/apply/deploy/verify it.

Use when the user asks "what can Cekura do", "what commands are available", "help me with Cekura", "what skills do I have", "show me Cekura features", "what's available", "how do I use Cekura", or needs guidance on which Cekura skill to use for their task. Also relevant as the entry point when a user has just installed cekura-skills for the first time.

Use when the user says "get started with Cekura", "set up Cekura", "onboard to Cekura", "I'm new to Cekura", "help me set up my agent", "how do I use Cekura", "walk me through Cekura", "configure my project", "first time using Cekura", or needs guidance on initial platform setup. Covers two onboarding paths: **testing** (default — build evaluators and run simulated calls) and **observability** (ingest production call logs and evaluate them).

Use when the user asks to "generate (test) scenarios", "generate evaluators", "create an evaluator", "create evals", "create a scenario", "write a test scenario", "design a test case", "test my agent", "build eval coverage", "plan a test suite", "create red team tests", "set up test profiles", "configure conditional actions", "build a deterministic test", "design an IVR test", "write a unit test for a voice agent", "build a regression test", "scripted scenario", "structured evaluator", or "run evals". Also for CHANGING existing evaluators — "update an evaluator", "improve my evals", "make these evaluators stricter", "add a DTMF step", "fix the expected outcome", "attach metrics to these" — and for debugging how the testing agent speaks: "why did it read the number as a word", "make it spell digits", "wrong language". Covers evaluator design and review, coverage, test profiles, mock-tool data, conditional actions, and red-team / edge-case practice.

Use when the user asks to "improve a metric", "run labs", "leave feedback on a metric", "add to labs", "fix metric accuracy", "review metric results", "find misaligned metrics", or "iterate on metric quality". Covers the metric improvement cycle, the feedback workflow, and the labs pipeline used to refine metric accuracy over time.

Use when the user asks to "create a new agent", "create a main agent", "set up a new agent", "set up a main agent", "add my main agent to Cekura", "configure my main agent", "connect my main agent", "set up mock tools", "add tools to my agent", "upload knowledge base", "configure integration", "connect VAPI", "connect Retell", "connect LiveKit", "connect ElevenLabs", "add dynamic variables", or needs to onboard a voice AI agent onto the Cekura platform. Covers the full agent setup flow: project selection, provider selection, basics and connection type, description, main agent creation, mock tools, knowledge base, dynamic variables, and advanced configuration.