← Quickchat AICONTENT HISTORY

Update to Quickchat AI

Snapshot Nov 9, 2026 · 18:08 UTC · version 2.0.7

Collection source: downloaded plugin package.

WHAT CHANGED

Instructions updated for launch-agent

Full description — Now says: run a confirmed customer question and show the actual reply.

Observed in instructions or declared skills. Runtime behavior has not been tested.

Save Quickchat AI and check future changes in Following.

Up to 3 plugins free · no card required.

All fields and original values.

Full description

Before

Oct 25, 2026 · 00:04 UTC

show the generated configuration and how to test it.

After

Nov 9, 2026 · 18:08 UTC

run a confirmed customer question and show the actual reply.

Skill instructions

Before

Oct 25, 2026 · 00:04 UTC

empty Agent the account already has (or provision one when every Agent is already configured), build it from a website URL or from a short interview when there is no site, then show the generated configuration and how to test it. U...

After

Nov 9, 2026 · 18:08 UTC

empty Agent the account already has (or provision one when every Agent is already configured), build it from a website URL or from a short interview when there is no site, then run a confirmed customer question and show the actual ...

Compare saved observations

Download comparison JSON

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -1,17 +1,17 @@
 ---
 name: launch-agent
 description: >-
-  Configure a Quickchat AI Agent from scratch, end to end: set up the empty Agent
-  the account already has (or provision one when every Agent is already
-  configured), build it from a website URL or from a short interview when there is
-  no site, then show the generated configuration and how to test it. Use
-  when the user asks to create, build, launch, set up, or spin up a new Agent, to
-  make an Agent from a website or URL, or when they have just connected Quickchat
-  and have nothing set up yet. Not for editing an established Agent (use
-  improve-agent).
+  Configure a Quickchat AI Agent from scratch, end to end: set up the empty
+  Agent the account already has (or provision one when every Agent is already
+  configured), build it from a website URL or from a short interview when there
+  is no site, then run a confirmed customer question and show the actual reply.
+  Use when the user asks to create, build, launch, set up, or spin up a new
+  Agent, to make an Agent from a website or URL, or when they have just
+  connected Quickchat and have nothing set up yet. Not for editing an
+  established Agent (use improve-agent).
 metadata:
   author: quickchat-ai
-  version: "2.2"
+  version: "2.3"
 ---
 
 # Launch a new Agent
@@ -19,6 +19,8 @@
 Stand up a working Agent in one guided flow. Most people running this have just
 connected Quickchat and already have an empty Agent, so configure THAT one and
 build something before you report anything.
+An OAuth connection alone is not a request to build or run a test. Start this
+flow when the user asks for it.
 
 ## Before you start
 1. Call `list_scenarios`. Connecting Quickchat already created an Agent, so if any
@@ -43,8 +45,11 @@
    content to any existing knowledge, so run it on an Agent that holds nothing
    yet; on an already-configured Agent, ask first.
 3. Poll `get_assistant_settings` until `onboarding_from_url_completed` is true.
+   Stop if `onboarding_from_url_blocked` appears; explain its `message` and ask
+   for a usable source instead of rerunning the rejected URL.
    If it reports `onboarding_progress_tracked` false there is no completion
-   signal — read the settings once after ~60s instead.
+   signal. Read the settings once after about 60 seconds and verify the persona
+   and knowledge before testing; do not claim a pending or failed build is ready.
 4. Show what was generated: the name, the main prompt, the guidelines and the
    language it picked.
 
@@ -85,12 +90,20 @@
 in the Quickchat dashboard (Actions page), then test it there.
 
 ## Close by making it real
-1. `get_deployment_info` — give them the widget embed snippet and the public chat
-   link.
-2. Tell them they can talk to it right now at
+1. Once setup is complete, ask for one customer question or suggest one grounded
+   in the knowledge just added. Explain that a test uses AI credits and can run
+   active AI Actions. Confirm the question and any action before testing.
+2. Call `send_message_to_agent` with the same scenario_id and the confirmed
+   question as `message`. Show the actual Agent reply here. A handoff placeholder,
+   allowance notice or error is not a successful AI answer; explain the result and resolve it
+   before claiming the Agent works. Do not invent an answer or silently retry a
+   call that may have run an Action.
+3. After an actual reply, offer one next step: improve that answer or share the
+   Agent. If they choose sharing, call `get_deployment_info` and give them the
+   widget snippet and public chat link. The preview remains available at
    `https://app.quickchat.ai/i/<scenario_id>/ai-preview`.
-3. Offer one concrete next step (add knowledge, tune the persona via
-   improve-agent).
+4. If the user declines testing or explicitly wants deployment, respect that
+   choice and provide the deployment information without claiming a test passed.
 
 ## Guardrails
 - Configure the Agent the account already has. Create at most ONE Agent per
@@ -105,4 +118,5 @@
 
 ## Output
 Confirm the Agent's name and scenario_id, summarize the configuration you set,
-and give the preview link plus 1-2 next steps.
+and show the actual test reply or clearly state why the test did not run or did
+not produce an AI answer. Offer one next step; do not end a requested test at a link.
Full technical diff · 2 changed fields

changed /description

BEFORE
"Configure a Quickchat AI Agent from scratch, end to end: set up the empty Agent the account already has (or provision one when every Agent is already configured), build it from a website URL or from a short interview when there is no site, then show the generated configuration and how to test it. Use when the user asks to create, build, launch, set up, or spin up a new Agent, to make an Agent from a website or URL, or when they have just connected Quickchat and have nothing set up yet. Not for editing an established Agent (use improve-agent)."
AFTER
"Configure a Quickchat AI Agent from scratch, end to end: set up the empty Agent the account already has (or provision one when every Agent is already configured), build it from a website URL or from a short interview when there is no site, then run a confirmed customer question and show the actual reply. Use when the user asks to create, build, launch, set up, or spin up a new Agent, to make an Agent from a website or URL, or when they have just connected Quickchat and have nothing set up yet. Not for editing an established Agent (use improve-agent)."

changed /skill_md_contents

BEFORE
"---\nname: launch-agent\ndescription: >-\n  Configure a Quickchat AI Agent from scratch, end to end: set up the empty Agent\n  the account already has (or provision one when every Agent is already\n  configured), build it from a website URL or from a short interview when there is\n  no site, then show the generated configuration and how to test it. Use\n  when the user asks to create, build, launch, set up, or spin up a new Agent, to\n  make an Agent from a website or URL, or when they have just connected Quickchat\n  and have nothing set up yet. Not for editing an established Agent (use\n  improve-agent).\nmetadata:\n  author: quickchat-ai\n  version: \"2.2\"\n---\n\n# Launch a new Agent\n\nStand up a working Agent in one guided flow. Most people running this have just\nconnected Quickchat and already have an empty Agent, so configure THAT one and\nbuild something before you report anything.\n\n## Before you start\n1. Call `list_scenarios`. Connecting Quickchat already created an Agent, so if any\n   entry has `configured: false` AND a `role` of EDITOR or above, use that\n   `scenario_id` for everything below and do NOT call `create_assistant` (creating\n   another leaves theirs blank forever). Ignore unconfigured Agents you only have\n   VIEWER on: they are someone else's and onboarding them will fail. Only when\n   every Agent you can edit is already configured, or the user asked for an extra\n   one, call `create_assistant` with confirm=true and use the scenario_id it\n   returns. If more than one Agent qualifies, ask which to set up.\n2. Ask whether they have a website to build from. That answer picks the path below.\n3. Their Agent is on the FREE tier (no charge, no card); say so if you create one.\n\n## Path A — they have a website\n1. Use the scenario_id from \"Before you start\". Set the display name with\n   `update_assistant_settings` if they gave one.\n2. Call `onboard_assistant_from_url` with that scenario_id, the URL and\n   confirm=true — all three are required, and the call is rejected without the\n   confirm. Say out loud that this takes 20-60 seconds while it scrapes the\n   site, writes the persona and embeds the content — silence here is where\n   people give up. It OVERWRITES persona and settings and adds the scraped\n   content to any existing knowledge, so run it on an Agent that holds nothing\n   yet; on an already-configured Agent, ask first.\n3. Poll `get_assistant_settings` until `onboarding_from_url_completed` is true.\n   If it reports `onboarding_progress_tracked` false there is no completion\n   signal — read the settings once after ~60s instead.\n4. Show what was generated: the name, the main prompt, the guidelines and the\n   language it picked.\n\nIf the URL is rejected as a placeholder or marketplace domain, ask for their own\nbusiness website rather than retrying the same URL.\n\n## Path B — no website\nDo NOT create an empty Agent and stop. Interview first, in one short round of\nquestions:\n- What does the business do, and what is it called?\n- Who will be talking to the Agent (customers, members, players, staff)?\n- What tone should it take?\n- What are the top 3 questions it must answer?\n\nThen configure their Agent in a single `update_assistant_settings` call on the\nscenario_id from \"Before you start\". Do this on every no-website run, including\nwhen you had to create the Agent: creation happens before this interview, so it\ncannot have carried answers you had not collected yet, and skipping the write\nleaves the new Agent blank.\n- `name` and `one_word_description` — what the Agent is called\n- `short_description` — the main prompt: role, what it helps with, what it must\n  not guess at\n- `ai_commands` — one short rule per item, from their must-answer questions\n- `greeting` — the first line a visitor sees\n- `language_chosen` — the language they answered in\n\nThen offer `add_knowledge_base_article` for any facts they can give you now\n(hours, pricing, policies). Content is embedded automatically; there is no\nretrain step.\n\n## Optional — connect a system\nIf they name something the Agent should reach (an order lookup, a booking system,\nan internal API), offer `create_http_request_action`, then\n`test_http_request_action` to prove it works. It takes no confirm argument. Skip\nthis unless they raise it, and never invent an endpoint.\nNever ask for an API key, token or password in this chat. If the API needs one,\ncreate the action without it and have the user add the credential to the action\nin the Quickchat dashboard (Actions page), then test it there.\n\n## Close by making it real\n1. `get_deployment_info` — give them the widget embed snippet and the public chat\n   link.\n2. Tell them they can talk to it right now at\n   `https://app.quickchat.ai/i/<scenario_id>/ai-preview`.\n3. Offer one concrete next step (add knowledge, tune the persona via\n   improve-agent).\n\n## Guardrails\n- Configure the Agent the account already has. Create at most ONE Agent per\n  request, only when none is unconfigured or the user asked; never loop\n  `create_assistant`. If its result carries a `warning` naming an Agent that\n  is still blank, tell the user and offer to configure that one instead.\n- If initial settings are rejected, the Agent still exists — adjust and apply with\n  `update_assistant_settings` on the returned scenario_id; do NOT call\n  `create_assistant` again.\n- If a configuration call returns an error, re-read with `get_assistant_settings`\n  before retrying; the write may already have landed.\n\n## Output\nConfirm the Agent's name and scenario_id, summarize the configuration you set,\nand give the preview link plus 1-2 next steps.\n"
AFTER
"---\nname: launch-agent\ndescription: >-\n  Configure a Quickchat AI Agent from scratch, end to end: set up the empty\n  Agent the account already has (or provision one when every Agent is already\n  configured), build it from a website URL or from a short interview when there\n  is no site, then run a confirmed customer question and show the actual reply.\n  Use when the user asks to create, build, launch, set up, or spin up a new\n  Agent, to make an Agent from a website or URL, or when they have just\n  connected Quickchat and have nothing set up yet. Not for editing an\n  established Agent (use improve-agent).\nmetadata:\n  author: quickchat-ai\n  version: \"2.3\"\n---\n\n# Launch a new Agent\n\nStand up a working Agent in one guided flow. Most people running this have just\nconnected Quickchat and already have an empty Agent, so configure THAT one and\nbuild something before you report anything.\nAn OAuth connection alone is not a request to build or run a test. Start this\nflow when the user asks for it.\n\n## Before you start\n1. Call `list_scenarios`. Connecting Quickchat already created an Agent, so if any\n   entry has `configured: false` AND a `role` of EDITOR or above, use that\n   `scenario_id` for everything below and do NOT call `create_assistant` (creating\n   another leaves theirs blank forever). Ignore unconfigured Agents you only have\n   VIEWER on: they are someone else's and onboarding them will fail. Only when\n   every Agent you can edit is already configured, or the user asked for an extra\n   one, call `create_assistant` with confirm=true and use the scenario_id it\n   returns. If more than one Agent qualifies, ask which to set up.\n2. Ask whether they have a website to build from. That answer picks the path below.\n3. Their Agent is on the FREE tier (no charge, no card); say so if you create one.\n\n## Path A — they have a website\n1. Use the scenario_id from \"Before you start\". Set the display name with\n   `update_assistant_settings` if they gave one.\n2. Call `onboard_assistant_from_url` with that scenario_id, the URL and\n   confirm=true — all three are required, and the call is rejected without the\n   confirm. Say out loud that this takes 20-60 seconds while it scrapes the\n   site, writes the persona and embeds the content — silence here is where\n   people give up. It OVERWRITES persona and settings and adds the scraped\n   content to any existing knowledge, so run it on an Agent that holds nothing\n   yet; on an already-configured Agent, ask first.\n3. Poll `get_assistant_settings` until `onboarding_from_url_completed` is true.\n   Stop if `onboarding_from_url_blocked` appears; explain its `message` and ask\n   for a usable source instead of rerunning the rejected URL.\n   If it reports `onboarding_progress_tracked` false there is no completion\n   signal. Read the settings once after about 60 seconds and verify the persona\n   and knowledge before testing; do not claim a pending or failed build is ready.\n4. Show what was generated: the name, the main prompt, the guidelines and the\n   language it picked.\n\nIf the URL is rejected as a placeholder or marketplace domain, ask for their own\nbusiness website rather than retrying the same URL.\n\n## Path B — no website\nDo NOT create an empty Agent and stop. Interview first, in one short round of\nquestions:\n- What does the business do, and what is it called?\n- Who will be talking to the Agent (customers, members, players, staff)?\n- What tone should it take?\n- What are the top 3 questions it must answer?\n\nThen configure their Agent in a single `update_assistant_settings` call on the\nscenario_id from \"Before you start\". Do this on every no-website run, including\nwhen you had to create the Agent: creation happens before this interview, so it\ncannot have carried answers you had not collected yet, and skipping the write\nleaves the new Agent blank.\n- `name` and `one_word_description` — what the Agent is called\n- `short_description` — the main prompt: role, what it helps with, what it must\n  not guess at\n- `ai_commands` — one short rule per item, from their must-answer questions\n- `greeting` — the first line a visitor sees\n- `language_chosen` — the language they answered in\n\nThen offer `add_knowledge_base_article` for any facts they can give you now\n(hours, pricing, policies). Content is embedded automatically; there is no\nretrain step.\n\n## Optional — connect a system\nIf they name something the Agent should reach (an order lookup, a booking system,\nan internal API), offer `create_http_request_action`, then\n`test_http_request_action` to prove it works. It takes no confirm argument. Skip\nthis unless they raise it, and never invent an endpoint.\nNever ask for an API key, token or password in this chat. If the API needs one,\ncreate the action without it and have the user add the credential to the action\nin the Quickchat dashboard (Actions page), then test it there.\n\n## Close by making it real\n1. Once setup is complete, ask for one customer question or suggest one grounded\n   in the knowledge just added. Explain that a test uses AI credits and can run\n   active AI Actions. Confirm the question and any action before testing.\n2. Call `send_message_to_agent` with the same scenario_id and the confirmed\n   question as `message`. Show the actual Agent reply here. A handoff placeholder,\n   allowance notice or error is not a successful AI answer; explain the result and resolve it\n   before claiming the Agent works. Do not invent an answer or silently retry a\n   call that may have run an Action.\n3. After an actual reply, offer one next step: improve that answer or share the\n   Agent. If they choose sharing, call `get_deployment_info` and give them the\n   widget snippet and public chat link. The preview remains available at\n   `https://app.quickchat.ai/i/<scenario_id>/ai-preview`.\n4. If the user declines testing or explicitly wants deployment, respect that\n   choice and provide the deployment information without claiming a test passed.\n\n## Guardrails\n- Configure the Agent the account already has. Create at most ONE Agent per\n  request, only when none is unconfigured or the user asked; never loop\n  `create_assistant`. If its result carries a `warning` naming an Agent that\n  is still blank, tell the user and offer to configure that one instead.\n- If initial settings are rejected, the Agent still exists — adjust and apply with\n  `update_assistant_settings` on the returned scenario_id; do NOT call\n  `create_assistant` again.\n- If a configuration call returns an error, re-read with `get_assistant_settings`\n  before retrying; the write may already have landed.\n\n## Output\nConfirm the Agent's name and scenario_id, summarize the configuration you set,\nand show the actual test reply or clearly state why the test did not run or did\nnot produce an AI answer. Offer one next step; do not end a requested test at a link.\n"
Full snapshot data
{
  "description": "Configure a Quickchat AI Agent from scratch, end to end: set up the empty Agent the account already has (or provision one when every Agent is already configured), build it from a website URL or from a short interview when there is no site, then run a confirmed customer question and show the actual reply. Use when the user asks to create, build, launch, set up, or spin up a new Agent, to make an Agent from a website or URL, or when they have just connected Quickchat and have nothing set up yet. Not for editing an established Agent (use improve-agent).",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 197
    }
  ],
  "name": "launch-agent",
  "skill_md_contents": "---\nname: launch-agent\ndescription: >-\n  Configure a Quickchat AI Agent from scratch, end to end: set up the empty\n  Agent the account already has (or provision one when every Agent is already\n  configured), build it from a website URL or from a short interview when there\n  is no site, then run a confirmed customer question and show the actual reply.\n  Use when the user asks to create, build, launch, set up, or spin up a new\n  Agent, to make an Agent from a website or URL, or when they have just\n  connected Quickchat and have nothing set up yet. Not for editing an\n  established Agent (use improve-agent).\nmetadata:\n  author: quickchat-ai\n  version: \"2.3\"\n---\n\n# Launch a new Agent\n\nStand up a working Agent in one guided flow. Most people running this have just\nconnected Quickchat and already have an empty Agent, so configure THAT one and\nbuild something before you report anything.\nAn OAuth connection alone is not a request to build or run a test. Start this\nflow when the user asks for it.\n\n## Before you start\n1. Call `list_scenarios`. Connecting Quickchat already created an Agent, so if any\n   entry has `configured: false` AND a `role` of EDITOR or above, use that\n   `scenario_id` for everything below and do NOT call `create_assistant` (creating\n   another leaves theirs blank forever). Ignore unconfigured Agents you only have\n   VIEWER on: they are someone else's and onboarding them will fail. Only when\n   every Agent you can edit is already configured, or the user asked for an extra\n   one, call `create_assistant` with confirm=true and use the scenario_id it\n   returns. If more than one Agent qualifies, ask which to set up.\n2. Ask whether they have a website to build from. That answer picks the path below.\n3. Their Agent is on the FREE tier (no charge, no card); say so if you create one.\n\n## Path A — they have a website\n1. Use the scenario_id from \"Before you start\". Set the display name with\n   `update_assistant_settings` if they gave one.\n2. Call `onboard_assistant_from_url` with that scenario_id, the URL and\n   confirm=true — all three are required, and the call is rejected without the\n   confirm. Say out loud that this takes 20-60 seconds while it scrapes the\n   site, writes the persona and embeds the content — silence here is where\n   people give up. It OVERWRITES persona and settings and adds the scraped\n   content to any existing knowledge, so run it on an Agent that holds nothing\n   yet; on an already-configured Agent, ask first.\n3. Poll `get_assistant_settings` until `onboarding_from_url_completed` is true.\n   Stop if `onboarding_from_url_blocked` appears; explain its `message` and ask\n   for a usable source instead of rerunning the rejected URL.\n   If it reports `onboarding_progress_tracked` false there is no completion\n   signal. Read the settings once after about 60 seconds and verify the persona\n   and knowledge before testing; do not claim a pending or failed build is ready.\n4. Show what was generated: the name, the main prompt, the guidelines and the\n   language it picked.\n\nIf the URL is rejected as a placeholder or marketplace domain, ask for their own\nbusiness website rather than retrying the same URL.\n\n## Path B — no website\nDo NOT create an empty Agent and stop. Interview first, in one short round of\nquestions:\n- What does the business do, and what is it called?\n- Who will be talking to the Agent (customers, members, players, staff)?\n- What tone should it take?\n- What are the top 3 questions it must answer?\n\nThen configure their Agent in a single `update_assistant_settings` call on the\nscenario_id from \"Before you start\". Do this on every no-website run, including\nwhen you had to create the Agent: creation happens before this interview, so it\ncannot have carried answers you had not collected yet, and skipping the write\nleaves the new Agent blank.\n- `name` and `one_word_description` — what the Agent is called\n- `short_description` — the main prompt: role, what it helps with, what it must\n  not guess at\n- `ai_commands` — one short rule per item, from their must-answer questions\n- `greeting` — the first line a visitor sees\n- `language_chosen` — the language they answered in\n\nThen offer `add_knowledge_base_article` for any facts they can give you now\n(hours, pricing, policies). Content is embedded automatically; there is no\nretrain step.\n\n## Optional — connect a system\nIf they name something the Agent should reach (an order lookup, a booking system,\nan internal API), offer `create_http_request_action`, then\n`test_http_request_action` to prove it works. It takes no confirm argument. Skip\nthis unless they raise it, and never invent an endpoint.\nNever ask for an API key, token or password in this chat. If the API needs one,\ncreate the action without it and have the user add the credential to the action\nin the Quickchat dashboard (Actions page), then test it there.\n\n## Close by making it real\n1. Once setup is complete, ask for one customer question or suggest one grounded\n   in the knowledge just added. Explain that a test uses AI credits and can run\n   active AI Actions. Confirm the question and any action before testing.\n2. Call `send_message_to_agent` with the same scenario_id and the confirmed\n   question as `message`. Show the actual Agent reply here. A handoff placeholder,\n   allowance notice or error is not a successful AI answer; explain the result and resolve it\n   before claiming the Agent works. Do not invent an answer or silently retry a\n   call that may have run an Action.\n3. After an actual reply, offer one next step: improve that answer or share the\n   Agent. If they choose sharing, call `get_deployment_info` and give them the\n   widget snippet and public chat link. The preview remains available at\n   `https://app.quickchat.ai/i/<scenario_id>/ai-preview`.\n4. If the user declines testing or explicitly wants deployment, respect that\n   choice and provide the deployment information without claiming a test passed.\n\n## Guardrails\n- Configure the Agent the account already has. Create at most ONE Agent per\n  request, only when none is unconfigured or the user asked; never loop\n  `create_assistant`. If its result carries a `warning` naming an Agent that\n  is still blank, tell the user and offer to configure that one instead.\n- If initial settings are rejected, the Agent still exists — adjust and apply with\n  `update_assistant_settings` on the returned scenario_id; do NOT call\n  `create_assistant` again.\n- If a configuration call returns an error, re-read with `get_assistant_settings`\n  before retrying; the write may already have landed.\n\n## Output\nConfirm the Agent's name and scenario_id, summarize the configuration you set,\nand show the actual test reply or clearly state why the test did not run or did\nnot produce an AI answer. Offer one next step; do not end a requested test at a link.\n"
}

SHA-256 of public snapshot: 1123d112012a532cc37197533b97db10cd042e0e47115adb2403b96f8f1f2855