← Quickchat AICONTENT HISTORY

Update to Quickchat AI

Snapshot Oct 25, 2026 · 00:04 UTC · version 2.0.6

Collection source: downloaded plugin package. These snapshots do not have a confirmed matching collection source. Differences in file lists alone do not establish changes to the package.

WHAT CHANGED

Instructions updated for improve-agent

Skill instructions — Now says: immediately, no retrain. `ai_commands` REPLACES every existing guideline, so read the current list first and send it back with the change. The list is ...

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.

Skill instructions

Before

Sep 30, 2026 · 22:49 UTC

immediately, no retrain.

After

Oct 25, 2026 · 00:04 UTC

immediately, no retrain. `ai_commands` REPLACES every existing guideline, so read the current list first and send it back with the change. The list is capped at 10,000 characters in total: near the cap, merge or shorten guideline...

Compare saved observations

Download comparison JSON

SKILL.md text changes

− removed · + added

--- before
+++ after
@@ -42,7 +42,10 @@
 
 ## Apply (only after approval)
 - `update_assistant_settings` — pass ONLY the fields that change; effective
-  immediately, no retrain.
+  immediately, no retrain. `ai_commands` REPLACES every existing guideline, so
+  read the current list first and send it back with the change. The list is
+  capped at 10,000 characters in total: near the cap, merge or shorten
+  guidelines rather than appending one more, which can only be rejected.
 - `add_knowledge_base_article` — content required; embedded automatically in the
   background, no retrain.
 - `update_knowledge_base_article` — replaces the article's WHOLE body, so read it
Full technical diff · 1 changed fields

changed /skill_md_contents

BEFORE
"---\nname: improve-agent\ndescription: >-\n  Audit a Quickchat AI Agent and improve it: read its current settings and\n  knowledge base, mine recent conversations for gaps (unanswered questions, wrong\n  answers, off-brand tone, missing knowledge), then propose specific\n  configuration edits and knowledge-base additions. Apply changes only after the\n  user approves each one. Use when the user asks to improve, tune, optimize, fix,\n  or update an Agent, add knowledge, or change its persona, greeting, language,\n  or profile picture.\nmetadata:\n  author: quickchat-ai\n  version: \"1.0\"\n---\n\n# Improve an Agent\n\nRead first, change second. Diagnose from real data, propose the smallest fix, and\nnever write without explicit approval.\n\n## Before you start\n- Call `list_scenarios` to resolve the Agent. Config tools require EDITOR-or-above\n  on that Agent; if the user lacks it, say so and stop.\n- Ask whether to PROPOSE changes only (default) or also APPLY approved ones.\n\n## Diagnose\n1. `get_assistant_settings` — persona, profession, creativity, greeting, language,\n   reply length, KB descriptions.\n2. `list_knowledge_base_articles` — what the Agent already knows. `content` here\n   is a preview, not the article; `get_knowledge_base_article` returns the body.\n3. `get_insights` plus a `list_conversations` sample read with\n   `get_conversation_detail` to find where it struggled. Classify with\n   `references/failure-taxonomy.md`.\n\n## Propose\nFor each recurring problem, propose the SMALLEST change that fixes it:\n- Missing knowledge -> a new KB article (draft the exact text).\n- Wrong tone/persona -> a specific `personality` (an enum id, not a scale) or an\n  `ai_commands` guideline.\n- Too long / short / robotic -> `reply_length` or `creativity_level`.\nPresent proposals as a before -> after diff, then STOP for approval.\n\n## Apply (only after approval)\n- `update_assistant_settings` — pass ONLY the fields that change; effective\n  immediately, no retrain.\n- `add_knowledge_base_article` — content required; embedded automatically in the\n  background, no retrain.\n- `update_knowledge_base_article` — replaces the article's WHOLE body, so read it\n  with `get_knowledge_base_article` first and send the full edited text.\n  `previous_title` + `previous_content` restore the previous title and\n  body, unless `previous_content_truncated` is true.\n- `delete_knowledge_base_article` — irreversible; read the article first, since\n  the response echoes only the first 20,000 characters for recovery. It requires\n  `expected_title` and `expected_added_at` from\n  `list_knowledge_base_articles` (they fail the call if the article changed\n  underneath you) plus confirm=true.\n- `set_assistant_avatar` — profile picture from an image URL, a base64 data URI\n  (png/jpg/gif/webp, max 5 MB), or `use_default=true` to reset. It permanently\n  replaces the previous image and the legacy launcher icon, and needs a plan\n  that allows widget customization. It returns the resulting `avatar_url` —\n  there is no settings field to re-read it from.\n- After each other write, re-read with `get_assistant_settings` /\n  `list_knowledge_base_articles` and confirm the change landed.\n\n## Guardrails\n- Never call a write tool before the user approves that specific change.\n- Never build an edit out of a preview — correcting one line still means sending\n  the whole article, and a preview sent back silently drops the rest.\n- `personality` and similar fields are enum ids — use the ids from the tool\n  schema, not free text.\n- Some fields (e.g. `ai_commands`) need a plan feature; if a write is rejected for\n  that reason, report it plainly instead of retrying.\n\n## Output\nA short report: top issues (ranked, one example each), the proposed changes as a\ndiff, and — if you applied any — a confirmation of what changed.\n"
AFTER
"---\nname: improve-agent\ndescription: >-\n  Audit a Quickchat AI Agent and improve it: read its current settings and\n  knowledge base, mine recent conversations for gaps (unanswered questions, wrong\n  answers, off-brand tone, missing knowledge), then propose specific\n  configuration edits and knowledge-base additions. Apply changes only after the\n  user approves each one. Use when the user asks to improve, tune, optimize, fix,\n  or update an Agent, add knowledge, or change its persona, greeting, language,\n  or profile picture.\nmetadata:\n  author: quickchat-ai\n  version: \"1.0\"\n---\n\n# Improve an Agent\n\nRead first, change second. Diagnose from real data, propose the smallest fix, and\nnever write without explicit approval.\n\n## Before you start\n- Call `list_scenarios` to resolve the Agent. Config tools require EDITOR-or-above\n  on that Agent; if the user lacks it, say so and stop.\n- Ask whether to PROPOSE changes only (default) or also APPLY approved ones.\n\n## Diagnose\n1. `get_assistant_settings` — persona, profession, creativity, greeting, language,\n   reply length, KB descriptions.\n2. `list_knowledge_base_articles` — what the Agent already knows. `content` here\n   is a preview, not the article; `get_knowledge_base_article` returns the body.\n3. `get_insights` plus a `list_conversations` sample read with\n   `get_conversation_detail` to find where it struggled. Classify with\n   `references/failure-taxonomy.md`.\n\n## Propose\nFor each recurring problem, propose the SMALLEST change that fixes it:\n- Missing knowledge -> a new KB article (draft the exact text).\n- Wrong tone/persona -> a specific `personality` (an enum id, not a scale) or an\n  `ai_commands` guideline.\n- Too long / short / robotic -> `reply_length` or `creativity_level`.\nPresent proposals as a before -> after diff, then STOP for approval.\n\n## Apply (only after approval)\n- `update_assistant_settings` — pass ONLY the fields that change; effective\n  immediately, no retrain. `ai_commands` REPLACES every existing guideline, so\n  read the current list first and send it back with the change. The list is\n  capped at 10,000 characters in total: near the cap, merge or shorten\n  guidelines rather than appending one more, which can only be rejected.\n- `add_knowledge_base_article` — content required; embedded automatically in the\n  background, no retrain.\n- `update_knowledge_base_article` — replaces the article's WHOLE body, so read it\n  with `get_knowledge_base_article` first and send the full edited text.\n  `previous_title` + `previous_content` restore the previous title and\n  body, unless `previous_content_truncated` is true.\n- `delete_knowledge_base_article` — irreversible; read the article first, since\n  the response echoes only the first 20,000 characters for recovery. It requires\n  `expected_title` and `expected_added_at` from\n  `list_knowledge_base_articles` (they fail the call if the article changed\n  underneath you) plus confirm=true.\n- `set_assistant_avatar` — profile picture from an image URL, a base64 data URI\n  (png/jpg/gif/webp, max 5 MB), or `use_default=true` to reset. It permanently\n  replaces the previous image and the legacy launcher icon, and needs a plan\n  that allows widget customization. It returns the resulting `avatar_url` —\n  there is no settings field to re-read it from.\n- After each other write, re-read with `get_assistant_settings` /\n  `list_knowledge_base_articles` and confirm the change landed.\n\n## Guardrails\n- Never call a write tool before the user approves that specific change.\n- Never build an edit out of a preview — correcting one line still means sending\n  the whole article, and a preview sent back silently drops the rest.\n- `personality` and similar fields are enum ids — use the ids from the tool\n  schema, not free text.\n- Some fields (e.g. `ai_commands`) need a plan feature; if a write is rejected for\n  that reason, report it plainly instead of retrying.\n\n## Output\nA short report: top issues (ranked, one example each), the proposed changes as a\ndiff, and — if you applied any — a confirmation of what changed.\n"
Full snapshot data
{
  "description": "Audit a Quickchat AI Agent and improve it: read its current settings and knowledge base, mine recent conversations for gaps (unanswered questions, wrong answers, off-brand tone, missing knowledge), then propose specific configuration edits and knowledge-base additions. Apply changes only after the user approves each one. Use when the user asks to improve, tune, optimize, fix, or update an Agent, add knowledge, or change its persona, greeting, language, or profile picture.",
  "included_files": [
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 193
    },
    {
      "relative_path": "references/failure-taxonomy.md",
      "size_in_bytes": 941
    }
  ],
  "name": "improve-agent",
  "skill_md_contents": "---\nname: improve-agent\ndescription: >-\n  Audit a Quickchat AI Agent and improve it: read its current settings and\n  knowledge base, mine recent conversations for gaps (unanswered questions, wrong\n  answers, off-brand tone, missing knowledge), then propose specific\n  configuration edits and knowledge-base additions. Apply changes only after the\n  user approves each one. Use when the user asks to improve, tune, optimize, fix,\n  or update an Agent, add knowledge, or change its persona, greeting, language,\n  or profile picture.\nmetadata:\n  author: quickchat-ai\n  version: \"1.0\"\n---\n\n# Improve an Agent\n\nRead first, change second. Diagnose from real data, propose the smallest fix, and\nnever write without explicit approval.\n\n## Before you start\n- Call `list_scenarios` to resolve the Agent. Config tools require EDITOR-or-above\n  on that Agent; if the user lacks it, say so and stop.\n- Ask whether to PROPOSE changes only (default) or also APPLY approved ones.\n\n## Diagnose\n1. `get_assistant_settings` — persona, profession, creativity, greeting, language,\n   reply length, KB descriptions.\n2. `list_knowledge_base_articles` — what the Agent already knows. `content` here\n   is a preview, not the article; `get_knowledge_base_article` returns the body.\n3. `get_insights` plus a `list_conversations` sample read with\n   `get_conversation_detail` to find where it struggled. Classify with\n   `references/failure-taxonomy.md`.\n\n## Propose\nFor each recurring problem, propose the SMALLEST change that fixes it:\n- Missing knowledge -> a new KB article (draft the exact text).\n- Wrong tone/persona -> a specific `personality` (an enum id, not a scale) or an\n  `ai_commands` guideline.\n- Too long / short / robotic -> `reply_length` or `creativity_level`.\nPresent proposals as a before -> after diff, then STOP for approval.\n\n## Apply (only after approval)\n- `update_assistant_settings` — pass ONLY the fields that change; effective\n  immediately, no retrain. `ai_commands` REPLACES every existing guideline, so\n  read the current list first and send it back with the change. The list is\n  capped at 10,000 characters in total: near the cap, merge or shorten\n  guidelines rather than appending one more, which can only be rejected.\n- `add_knowledge_base_article` — content required; embedded automatically in the\n  background, no retrain.\n- `update_knowledge_base_article` — replaces the article's WHOLE body, so read it\n  with `get_knowledge_base_article` first and send the full edited text.\n  `previous_title` + `previous_content` restore the previous title and\n  body, unless `previous_content_truncated` is true.\n- `delete_knowledge_base_article` — irreversible; read the article first, since\n  the response echoes only the first 20,000 characters for recovery. It requires\n  `expected_title` and `expected_added_at` from\n  `list_knowledge_base_articles` (they fail the call if the article changed\n  underneath you) plus confirm=true.\n- `set_assistant_avatar` — profile picture from an image URL, a base64 data URI\n  (png/jpg/gif/webp, max 5 MB), or `use_default=true` to reset. It permanently\n  replaces the previous image and the legacy launcher icon, and needs a plan\n  that allows widget customization. It returns the resulting `avatar_url` —\n  there is no settings field to re-read it from.\n- After each other write, re-read with `get_assistant_settings` /\n  `list_knowledge_base_articles` and confirm the change landed.\n\n## Guardrails\n- Never call a write tool before the user approves that specific change.\n- Never build an edit out of a preview — correcting one line still means sending\n  the whole article, and a preview sent back silently drops the rest.\n- `personality` and similar fields are enum ids — use the ids from the tool\n  schema, not free text.\n- Some fields (e.g. `ai_commands`) need a plan feature; if a write is rejected for\n  that reason, report it plainly instead of retrying.\n\n## Output\nA short report: top issues (ranked, one example each), the proposed changes as a\ndiff, and — if you applied any — a confirmation of what changed.\n"
}

SHA-256 of public snapshot: 1b3c6b461cb55084af0d2f79c653e80fc9fb5e227f9df4298ed795c7337e53b6