Skip to content

tough-tongue/toughtongue-skills / skills/ttai-agent

ttai-agent

Verified

Creates, edits, and refines Tough Tongue AI Scenarios, the definitions behind voice agents (sales calls, lead qualification, demos, screening, support, booking) and practice roleplays (negotiations, interviews, coaching). Writes ai_instructions, ai_model_config, strategy, tools_config, and rubrik through ttai:create_scenario and ttai:update_scenario. Also resolves the personal or organization workspace, checks plan and balance, places or cancels SIP calls, schedules or cancels meeting bots, mints private Scenario links, grounds a Scenario in the user's documents with knowledge FileBases, and troubleshoots ttai MCP sign-in. Use when the user wants a voice agent or practice Scenario created or improved ("create a voice agent", "the agent ends calls too early", "make it sound more natural"), a call placed, a meeting bot sent, or a Tough Tongue AI account question answered. Not for reports across many sessions (ttai-session-analyst) or browser demo steps (ttai-browser-demo-builder).

Tough Tongue AI Agent

Tough Tongue AI is the platform for tough conversations. Some the AI takes: voice agents that call, demo, screen, support, and book. Others you nail: realistic roleplay for negotiations, interviews, sales, and coaching.

This skill turns a person's intent into the smallest correct action on their Tough Tongue AI workspace through the ttai MCP server. Live tool schemas and tool results always win over this text.

Contents

  • Mental model
  • Scenario workflow (create / edit / refine)
  • Job router
  • Hard rules
  • Key Files

Mental model

  • Workspace — a personal account or an organization (org_id from ttai:get_workspace_info organizations).
  • Scenario — one agent definition: instructions, model stamp, controls, rubric.
  • Run — a session on the web link, an embed, a SIP call, or a meeting bot.
  • Result — transcript, evaluation, extraction; read per session or in aggregate.
  • Access — server-decided role, plan, and balance; never inferred from a Scenario.

Scenario workflow

The core job. Full procedure: references/scenario/workflow.md.

  • Create — the user has a new outcome, brief, or transcript. Finish with ttai:create_scenario (no id).
  • Edit — the user has a known Scenario and a requested change. Finish with ttai:update_scenario (id + changed fields).
  • Refine — the user has a complaint or a poor session. Finish with evidence → one diagnosis → surgical update.

Loop: resolve workspace → fetch current Scenario (ttai:get_resource(scenarios)) → load the matching recipe and Scenario references → draft → load the live write schema → write → re-fetch → report the change, the run link https://app.toughtongueai.com/run/<scenario_id>, and that edits affect new sessions only.

Job router

Read only the file the job needs. Each file is self-contained. Each entry gives the job, the file, and when to read it.

Account, access, and runs

  • Scope, focus, first-run orientation — operating-model.md. Any stateful request; "what can I do here?"
  • What a workspace holds — entities/resources.md. Listing or explaining resources.
  • Workspaces, plans, balance, sharing — entities/account-and-access.md. Org vs personal, entitlements, private links, SATs.
  • Run a Scenario, read results — entities/scenario-engine.md. Web link, embed, SIP, meeting bot, calling campaigns (office hours, retries, voicemail), session results.

Scenario authoring

  • Create / edit / refine a Scenario — scenario/workflow.md. Any Scenario write.
  • Design a good Scenario — scenario/authoring.md. Before drafting content.
  • Write ai_instructions — scenario/ai-instructions.md. Prompt shape, opening, spoken-line rules.
  • Personalize or branch a Scenario with variables — scenario/dynamic-variables.md. {{ var }} defaults, case/if variants, allow-lists, per-run facts.
  • Pick ai_model_config — scenario/model-selection.md. Model family, pipeline, voice.
  • Set controls — scenario/control.md. strategy, appearance, tools_config, analysis, links. A Google Slides deck (embed link, SLIDE MAP, slides instead of cards): its Google Slides decks section.
  • Answer from the user's documents — mcp/tools.md. PDFs or docs → knowledge FileBase → knowledge_filebase_ids.
  • Write the rubric — scenario/rubric.md. rubrik, processed_rubrik, who is scored. A rubrik always goes with session_analysis.is_auto_analysis: true; without it the rubric never runs.
  • Diagnose runtime behavior — scenario/runtime.md. Hang-ups, silence, wrap-up timing, tools over SIP.

Recipes

MCP

  • MCP tools, read policy, confirmations — mcp/tools.md. Choosing or calling any ttai: tool.
  • Connect a client, fix auth — mcp/clients.md. Install, OAuth, 401, missing tools.

No matching recipe (support, intake, survey, negotiation counterpart)? Use authoring + ai-instructions + control, and borrow the closest recipe's shape.

Sibling skills:

  • ttai-session-analyst — population-wide trends, scorecards, team reports.
  • ttai-browser-demo-builder — recorded browser steps and selectors for demos.
  • ttai-agent-apps — Agent Desktop apps the voice agent shows and edits.

Hard rules

  • MCP tools are the contract. Call only ttai: tools listed in mcp/tools.md and present in the live tools/list.
  • At the first Tough Tongue AI action in a conversation, call ttai:read_guide unless its guide is already in context.
  • Resolve workspace before a stateful call; pass the returned opaque org_id for organization work. Ask only when a write's scope is ambiguous.
  • Load the write schema before writing (Scenarios: ttai:get_schema("scenario")). Never invent fields, model IDs, voice IDs, or linked-resource IDs.
  • ttai:update_scenario replaces what you send. A tool's tool_settings and the whole ai_model_config are replaced, not merged: read, merge, write the full object.
  • Confirm the exact target (phone number, meeting URL, Scenario ID, audience) before any call, bot, delete, share, or post-processing.
  • Create Scenarios public: send is_public: true. Make one private only when the user asks to limit who can open it. Phone calls run private Scenarios too; a meeting bot runs a private one only with session_credentials: "bot_uc".
  • Separate what the platform supports from what this user's plan allows; the server decides. Treat 403 and balance errors as real blockers.
  • Advice stops at a recommendation. Never turn it into a write, call, or bot without the user's request.
  • Never ask for, paste, read, or log a credential; sign-in is the client's job. For CI or server setups, link headless setup.
  • Scenario edits affect new sessions only.

Key Files

reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexUnverifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall