Skip to content

fix: post_conversation human_id bug + prompt improvement#353

Draft
deepmasq wants to merge 2 commits intomainfrom
feat/karen-post-conversation-fixes
Draft

fix: post_conversation human_id bug + prompt improvement#353
deepmasq wants to merge 2 commits intomainfrom
feat/karen-post-conversation-fixes

Conversation

@deepmasq
Copy link
Copy Markdown
Contributor

Summary

  • Bug fix (#2426): bot_kanban_post_into_inbox() was called without human_id= param when spawning post_conversation tasks. The ktask_human_id column stayed empty, causing kanban safe scope to fail with "cannot determine human_id". Fixed in both karen_bot.py and marco_bot.py.
  • Prompt improvement (#2427): Restructured KAREN_POST_CONVERSATION from a terse 5-step list to explicit 4-round structure with parallel call hints. Inlined activity logging fields (saves a flexus_fetch_skill round). Added correct BANT scoring format and deal update instructions.
  • Scenario fix: Added missing fake_connected_providers: [] to post_conversation__bant_purchase.yaml.

Changes

File Change
karen_bot.py +1 line: human_id=new_task.ktask_human_id
marco_bot.py +1 line: same fix
karen_prompts.py Rewrite KAREN_POST_CONVERSATION (36 lines replacing 20)
post_conversation__bant_purchase.yaml +2 lines: fake_connected_providers: []

Test plan

  • Run post_conversation__bant_purchase scenario on staging
  • Verify kanban safe scope no longer errors on post_conversation tasks
  • Verify model completes in ~4 tool call rounds instead of 8+

🤖 Generated with Claude Code

Art Koval and others added 2 commits April 25, 2026 14:09
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…-based structure

Bug: bot_kanban_post_into_inbox() was called without human_id= param,
so ktask_human_id column stayed empty on post_conversation tasks.
The kanban safe scope then failed with "cannot determine human_id".

Prompt: restructured to 4 explicit rounds with parallel call hints,
inlined activity logging fields (saves a flexus_fetch_skill round),
added correct BANT scoring format and deal update instructions.

Also fixes same bug in marco_bot.py and adds missing
fake_connected_providers to post_conversation scenario YAML.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant