Ciarem AI

View saved version →

---
name: contact-brief
description: Use when the user wants to prepare for a call or meeting with a customer or lead, or asks "who is X" / "what do we know about X" in Ciarem.
---

# Contact brief

1. Call `search_contacts` with the name, phone or email the user gave. If several match, list them (name, phone, email) and ask which one.
2. Call `get_contact` with that `id`.
3. Call `get_conversation` on the newest of its `recent_conversations` (limit 30).
4. Write the brief:
   - **Who**: name, channels they use, city/country, where they came from (`origin`), funnel stage.
   - **Last conversation**: date, what they asked for, what we answered, anything left open.
   - **Open questions**: things they asked that nobody answered, promises made, objections.
   - **Suggested next step**: one sentence.
5. End with the contact's `url` so the user can open it in Ciarem.

Only state what the tools returned. If a field is empty, leave it out rather than guessing.

View saved version →

---
name: inbox-triage
description: Use when the user asks what needs their attention in Ciarem, which customers are waiting, or wants help catching up on their inbox.
---

# Inbox triage

1. Call `list_conversations` with `status: "open"` and `assigned_to_me: true` (limit 25). If the user is an admin and asks about the whole team, also call it with `unassigned: true`.
2. Rank by who has waited longest: `last_inbound_at` older than `last_message_at` of our reply, oldest first, then `unread_count`.
3. For the top 5, call `get_conversation` (limit 15) and write one line each: who, channel, what they want, how long they've waited.
4. Under each line, draft a short reply in the customer's language and tone. Say it is a draft: Ciarem's tools can't send messages from here. Link the conversation's `url` so the user can send it from Ciarem.
5. Skip conversations `handled_by: "ai"` unless the customer asked for a human; mention how many the AI is covering.

Never invent order numbers, prices or policies in a draft. If the answer needs a fact, call `search_knowledge` first; if it isn't there, say the draft needs that fact filled in.

View saved version →

---
name: set-up-workspace
description: Use when get_profile says `setup_required`, or a new Ciarem user wants to get started, set up their AI agent, or asks what to do first.
---

# Set up a new workspace

Admins only: if a tool answers `requires_admin`, tell the user a workspace admin has to do the setup.

1. **Ask the onboarding questions**, one or two at a time, in the user's language: company name, industry, what they offer, their website (optional but it makes the AI much better), what the AI should help with (`objectives`), and which channels they will use. Don't guess an answer.
2. **Call `set_up_workspace`.** Write one `agents` entry per kind the objectives give (the tool description lists the mapping), with a short name and when-to-use in the language the AI should answer in. If it answers `invalid_arguments` naming the kinds, call again with exactly those.
3. **Wait for the website.** While `get_setup_status` shows `website.ready: false`, tell the user it takes a few minutes, ask how they want the AI to sound (`update_ai_settings`: tone, reply length, welcome message), then check again. If `website.status` is `ERROR`, the site couldn't be read: say so and add the key facts (prices, hours, policies) with `add_knowledge_text` instead.
4. **Show it working.** Call `test_ai_reply` with two or three messages a real customer of theirs would send (a price question, a hours or location question). If `knowledge_still_indexing` isn't empty, wait and test again. Fix what's off the way the `tune-ai-agent` skill does.
5. **Finish.** When the user likes the answers, call `finish_setup` and share each `connect_channels` link: connecting needs their Meta login (or their website for web chat) in the Ciarem web app, and it is the last step before the AI answers real customers.

If any tool answers `subscription_required`, stop and share the payment link it gives.

View saved version →

---
name: tune-ai-agent
description: Use when an admin wants to set up, fix or improve their Ciarem AI agent, or says it answers something wrong, too long, too cold, or transfers when it shouldn't.
---

# Tune the AI agent

Admins only: if a tool answers `requires_admin`, tell the user only workspace admins can change the AI agent.

1. **Find what to fix.** Call `get_ai_setup`. If the user named no problem, also call `get_account_health`, `get_knowledge_gaps` and `get_ai_performance`, and propose the top three fixes. If they named a real conversation, call `explain_ai_decisions` for it.
2. **Reproduce it.** Call `test_ai_reply` with a customer message that shows the problem, and show the reply before changing anything.
3. **Pick the right lever**, one per problem:
   - Tone, language, length, initiative, welcome: `update_ai_settings`.
   - Who the business is and what it sells: `update_business_profile`.
   - A missing or wrong fact (price, hours, policy): `add_knowledge_text` into a knowledge base the answering agent uses (`update_agent` `knowledge_base_ids`), or `delete_knowledge_source` for a wrong one.
   - How to behave in a situation: `add_coaching`, with the customer message and the AI reply from step 2.
   - When to hand over to a person: `create_transfer_rule` / `update_transfer_rule`.
   - Which agent takes which conversations: `update_agent` `when_to_use`.
4. **Apply it.** Say what you will change in one line, then call the tool. Change only what the user asked for.
5. **Test it again.** Call `test_ai_reply` with the same message (a new `test_conversation_id` unless the problem needs the earlier turns), and compare before and after. If it isn't right yet, go back to step 3 with a different lever rather than piling on more corrections.
6. **Check for regressions** after a round of changes: `run_ai_checks`, wait a few minutes, then `get_ai_check_results` and report what didn't pass.
7. Close any recommended action the user fixed with `close_recommendation` (`resolve`).

Never turn the AI off (`set_ai_active`) or delete anything unless the user asked for it.

View saved version →

---
name: weekly-report
description: Use when an admin asks how the team, the inbox or the AI agent performed over the last week or day in Ciarem.
---

# Weekly report

Admins only: if a tool answers `requires_admin`, tell the user only workspace admins can see these numbers.

1. Call `get_metrics` with `period: "7d"` (or `"24h"` if the user asks about today).
2. Call `get_account_health`.
3. Write the report:
   - **Volume**: active conversations, per channel, new leads.
   - **AI agent**: how many conversations the AI handled, share of the total.
   - **Team**: the leaderboard as a small table (assigned, sent, resolved, median reply time in minutes).
   - **Waiting now**: `awaiting_reply`.
   - **Fix first**: the failing readiness checks and open recommended actions from `get_account_health`, most severe first.
4. If the user asks why the AI transferred or answered a specific chat, call `explain_ai_decisions` with the contact's name, phone or email.

Report the numbers as returned; don't extrapolate trends from a single period.