Skill Codex
Plan, execute, and verify Kore.ai Agent Platform work through the Arch MCP server using code-backed operation and dependency knowledge.
not rated 4
changed 1mo ago A 27 tokens
original MIT
MCP server Claude CodeCodexCursor +2
Arch MCP tools for Kore.ai Agent Platform (build, evaluate, optimize, debug, analyze). Runs locally from the @koreai/arch-mcp-tools npm package.
not rated 4 1mo ago A
tokens not measured
original MIT
Skill Codex
Start here for any question about building, debugging, testing, optimizing or shipping an agent on the Kore.ai Agent Platform with the Arch MCP tools, such as "why did my agent do that", "which sessions failed", "which topology should I use", "why is it slow or expensive", "how do I test this", "is voice set up right"…
not rated 4 1mo ago A 108 tokens
original MIT
Skill Codex
Explain and control what the model sees each turn on the Kore.ai Agent Platform, including system prompt sections, conversation history, session memory, tool results, knowledge-base results, compaction, truncation and secrets in context. Use when the model loses data it was given, repeats itself, when prompts or…
not rated 4 1mo ago A 82 tokens
original MIT
Skill Codex
Build and run evaluations for agents on the Kore.ai Agent Platform, including scenario catalogs, personas, evaluators, eval sets, eval runs and run-to-run comparison. Use when asked how to test an agent, to create or review test cases or evals, to measure whether a change helped, or when a project has no comprehensive…
not rated 4 1mo ago A 91 tokens
original MIT
Skill Codex
Define what a good agent experience means before or early in a build on the Kore.ai Agent Platform. Use when starting a project, when asked for success criteria, latency or cost targets, model-call or token budgets, or when "optimize" has no measurable target. Produces a per-intent experience contract with success…
not rated 4 1mo ago A 87 tokens
original MIT
Skill Codex
Explain what happened in one Agent Platform session, turn, handoff, tool call, knowledge-base answer or guardrail decision, or find which sessions failed and why, in plain language. Use whenever asked to analyze, debug, diagnose or explain why an agent did something, missed an input, handed off, looped, went silent on…
not rated 4 1mo ago A 104 tokens
original MIT
Skill Codex
Set up, tune, test and debug knowledge bases (SearchAI) used by ABL agents on the Kore.ai Agent Platform, including ingestion and chunking, search tool bindings, retrieval settings (maxsearchtokens, Top K, similarity threshold, keyword weight, object-aware retrieval), permissions, citations, the automatic first…
not rated 4 1mo ago A 103 tokens
original MIT
Skill Codex
Reduce latency, model calls per turn, tokens and cost for agents on the Kore.ai Agent Platform. Use when an agent is slow, makes too many LLM calls, has a high token bill, has long time to first response or first audio, or when asked to optimize. Finds the most expensive turns, shows where each call and token goes…
not rated 4 1mo ago A 91 tokens
original MIT
Skill Codex
Check that a Kore.ai Agent Platform project is ready to publish or deploy. Use before a release, publish, deployment or customer demo, or when asked "is it ready to ship". Runs ABL lint and package validation, compares the eval set with the last release, and walks a pre-ship checklist for context, secrets, handoffs…
not rated 4 1mo ago A 82 tokens
original MIT
Skill Codex
Design, review and debug tools used by ABL agents on the Kore.ai Agent Platform, including HTTP, MCP, workflow, integration, knowledge-base and table tools; parameters and hidden runtime inputs; effects, confirmation and idempotency; result shaping with MODELCONTEXT and onresult; preconditions; errors and retries…
not rated 4 1mo ago A 112 tokens
original MIT
Skill Codex
Choose or review the agent topology of a Kore.ai Agent Platform project, including single agent vs scripted FLOW vs supervisor with specialists, handoff, delegate, fan-out, escalation, handoff context and multi-intent requests. Use when asked "which topology", "should I add an agent or specialist", "why does it route…
not rated 4 1mo ago A 90 tokens
original MIT
Skill Codex
Design and configure voice agents on the Kore.ai Agent Platform, for pipeline voice (speech-to-text, model, text-to-speech) and realtime speech-to-speech. Use for design and configuration questions about ASR quality and test sets, language switching, end-of-speech and silence timeouts, barge-in and greetings, fillers…
not rated 4 1mo ago A 109 tokens
original MIT
Skill Codex
Design, publish, call and debug workflows on the Kore.ai Agent Platform, including workflow-backed agent tools, sync vs async calls, timeouts, input and output mapping, versions (publish and activate), executions, step approvals and human tasks, triggers, retries and failure branches. Use when an agent calls a…
not rated 4 1mo ago A 89 tokens
original MIT
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: