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).

references/scenario/rubric.md

View on GitHub
# Rubric

How a session is evaluated. The API spells the fields `rubrik` (free-form text
you write) and `processed_rubrik` (structured criteria derived from it).

## Contents

- A rubric needs analysis on
- `rubrik`
- `processed_rubrik`
- Reading scores
- Agent rules

## A rubric needs analysis on

A `rubrik` is applied only when `session_analysis.is_auto_analysis` is true,
and the server default is `false`. A Scenario with a rubric and analysis off
shows the rubric in the web app but never scores a session, so the user gets no
report.

- **Create:** send `session_analysis` with `is_auto_analysis: true` in the same
  `ttai:create_scenario` call as `rubrik`.
- **Edit:** before adding or changing `rubrik`, read the current
  `session_analysis.is_auto_analysis`. If it is false or missing, set it to true
  in the same `ttai:update_scenario` call and tell the user.
- **Review or refine:** a Scenario that has a rubric with analysis off is a
  defect. Name it and offer to turn analysis on.

## `rubrik`

Author-facing text describing what "good" looks like and what the report must
contain. Recipes carry a starting pattern per situation (lead screening, rep
scorecard, demo intelligence, candidate evidence).

Declare the subject first: the human participant, the AI's counterpart, or
business facts extracted from the conversation. In multi-party sessions also set
`session_analysis.evaluation_target` (for example `"Sales Rep"`).

## `processed_rubrik`

Weighted criteria the evaluation pipeline scores against:

| Field                     | Meaning                                               |
| ------------------------- | ----------------------------------------------------- |
| `criteria[]`              | `topic`, `weight` (0–100), `prompt`, `requires_video` |
| `report_instructions`     | How to write the overall report                       |
| `next_steps_instructions` | How to write next steps                               |

The platform derives it from `rubrik`, and people refine it in the web app; a
manual edit is treated as intentional. Prefer writing clear `rubrik` text. Send
`processed_rubrik` only when the user asks for exact criteria, and then send the
whole object: it is replaced, not merged.

## Reading scores

Analysis runs only when `session_analysis.is_auto_analysis` is true, with or
without a `rubrik`.

- Compact: `ttai:list_resources(sessions)` with `include_fields: ["evaluation"]`.
- Full detail: `ttai:get_resource(sessions)` (`evaluation_results.final_score`,
  `evaluation_results.report_card`).
- Aggregates: `ttai:get_workspace_info(sections: [usage])`.
- A rubric change applies to sessions scored afterwards. Past scored sessions
  keep their results: `ttai:analyze_session` only fills missing results and
  never re-scores. Test a new rubric with a fresh session.

## Agent rules

1. Name who is evaluated; keep it consistent with ROLE, FLOW outcomes, and
   `user_instructions`.
2. Score observable stages, not vague traits. Weight the moments that matter.
3. Keep learning feedback (how the human did) separate from business
   intelligence (what happened).
4. Never claim criteria were reprocessed unless a tool result shows it.
5. When the scenario judges spoken answers by meaning, say so in `rubrik`: own
   words, a different order, mixed language or an imperfect transcript count
   when the idea lands. Name any items that must be exact. A report that grades
   wording contradicts a coach that judged meaning.

## Key Files

- [authoring.md](authoring.md) — evaluation principles
- [control.md](control.md) — `session_analysis` settings
- [ai-instructions.md](ai-instructions.md) — FLOW outcomes the rubric mirrors
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 CodeCodexVerifiedInstall
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: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated 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: 323kUpdated Claude CodeCodexVerifiedInstall