tough-tongue/toughtongue-skills / skills/ttai-agent
ttai-agent
VerifiedCreates, 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).
references/operating-model.md
View on GitHub# Operating model
How to turn a person's request into a scoped, evidenced action on their Tough
Tongue AI workspace. The same protocol serves a coding agent calling `ttai:`
tools and a web chat host that injects verified context and delegates tool calls
to its own adapter. Never assume a terminal, browser, filesystem, or a
particular model provider.
## Contents
- Consumer contract and context packet
- Context and capability protocol
- Account facts available through MCP
- First-run orientation
- Scope and focus
- Operation loop
## Consumer contract and context packet
Each input maps to one output:
- **Human request** → intent: orient, inspect, author, change, run, analyze, or
share.
- **Verified account / workspace context** → the selected personal or
organization scope.
- **Named resource or active screen** → the focused entity and its ID.
- **Consumer capabilities** → a safe tool/action plan.
- **Live tool result** → an evidence-backed next step or a precise blocker.
A host may hand over only verified, relevant facts in this optional shape. It is
a handoff convention, not a Tough Tongue AI resource or an authorization:
```json
{
"intent": "author | inspect | run | analyze",
"scope": { "kind": "organization", "org_id": "verified-id" },
"focus": { "kind": "scenario", "id": "verified-id" },
"account": { "source": "authenticated-host", "observed_at": "ISO-8601" },
"consumer": { "can_call_ttai": true, "can_render_actions": true }
}
```
Personal scope has no `org_id`. Focus may be `scenario`, `session`, or `none`.
Omit unknown fields. A web host may add a verified profile or entitlement
summary under `account`; an MCP-only consumer leaves it out. Never reuse a
packet across users, accounts, or organizations.
## Context and capability protocol
Reuse verified context from the same user and scope; do not re-fetch each turn.
At the first Tough Tongue AI action in a conversation, call
`ttai:read_guide` unless its guide is already in
context.
1. **Resolve scope.** No current scope → `ttai:get_workspace_info`. Pass an
opaque `organizations[].id` as `org_id` (never a slug); omit it for personal. Ask
only when personal vs organization is materially ambiguous for a write.
2. **Resolve focus.** A Scenario ID is authoritative. A title → one
`ttai:list_resources(scenarios, query: "<title>")`, then `ttai:get_resource(scenarios)`.
Account-level requests have no focus.
3. **Separate four facts.**
- **Supported** — Tough Tongue AI has the entity or operation.
- **Available** — this consumer can call the required tool.
- **Allowed** — the user's role, plan, and balance permit it.
- **Requested** — the user actually asked for it.
4. **Load evidence.** Read the target entity, the relevant reference, and the
live tool schema before a mutation.
5. **Act and explain.** State scope, changed entity, result, and any limitation,
without exposing private fields or credentials.
The server is authoritative for entitlement. A `403` or balance error is a real
blocker, not a cue to retry with invented parameters.
## Account facts available through MCP
All come from `ttai:get_workspace_info` sections:
- **User profile** — `me`. Do not infer identity beyond it.
- **Organization memberships** — `organizations`. Do not infer permissions
beyond the returned role.
- **Minutes balance** — `balance` (org wallet + your quota when quotas are on).
Do not infer whether a balance grants a feature.
- **Effective capabilities** — `plan`. Do not infer price, payment provider, or
billing history.
- **Members, usage** — `members` (org only), `usage`.
If `ttai:get_workspace_info` is absent from `tools/list`, say the deployment
does not publish that fact. Never fabricate account data in an MCP-only client.
## First-run orientation
For "get started", "is MCP working?", or "what can I do?", stay read-only:
1. Call `ttai:read_guide`. Missing tools or `401` →
[mcp/clients.md](mcp/clients.md); never ask for a token in chat.
2. Call `ttai:get_workspace_info`.
3. Take a light inventory: `ttai:list_resources(scenarios)` and `ttai:list_resources(sessions)`
with small limits (`include_total: true` only when the count matters). Report
only what the responses show.
4. Route to the next job:
- New goal or no Scenario → create
([scenario/workflow.md](scenario/workflow.md)).
- Existing Scenario to change or fix → edit or refine (same file).
- Session evidence across a team → the `ttai-session-analyst` skill.
- Browser walkthrough to record → the `ttai-browser-demo-builder` skill.
If the person already named a job, skip inventory. Never create, update, call,
schedule, or share during orientation without a separate request.
## Scope and focus
A **workspace** is personal or an organization. A **focus** is the resource
under discussion — a conversation concept, not a stored resource.
Each focus lists its first read, then the usual next action:
- **Scenario** — known ID or one title search → create, update, share, or run.
- **Sessions for a Scenario** — `ttai:list_resources(sessions)` → read evidence or
analyze.
- **SIP call** — `ttai:list_resources(sip-trunks)`, `ttai:list_resources(bots)` → place or
inspect a call.
- **Meeting bot** — `ttai:list_resources(bots)` → schedule or inspect.
- **Account / workspace** — organizations and small typed lists → recommend a
workflow.
Do not change a focused Scenario while the request is only about learning from
it. Session evidence can justify a later, separate refinement.
## Operation loop
1. **Understand** — restate the outcome and the operation type.
2. **Ground** — resolve scope, focus, permissions, and facts from live data.
3. **Choose** — the narrowest resource and channel that meets the goal.
4. **Prepare** — load the relevant reference and the live tool schema.
5. **Execute** — the smallest requested change; confirm external or destructive
actions first ([mcp/tools.md](mcp/tools.md#confirm-before-acting)).
6. **Verify** — re-read what changed. Sessions and post-processing are async.
7. **Hand back** — separate facts, recommendation, action taken, and open
decisions.
Advice-only requests stop after **Choose**.
## Key Files
- [mcp/tools.md](mcp/tools.md) — tool catalog, read policy, confirmations
- [entities/account-and-access.md](entities/account-and-access.md) — workspaces,
plans, sharing
- [scenario/workflow.md](scenario/workflow.md) — Scenario writes
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| 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: 472k | Updated | Claude CodeCodex | Verified | Install |
| 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: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| 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 available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |