Skip to content

CMP-229: Strengthen critical instructions with persuasion principles#49

Open
mal-leanix wants to merge 2 commits intomainfrom
feature/cmp-229-persuasion-principles
Open

CMP-229: Strengthen critical instructions with persuasion principles#49
mal-leanix wants to merge 2 commits intomainfrom
feature/cmp-229-persuasion-principles

Conversation

@mal-leanix
Copy link
Contributor

@mal-leanix mal-leanix commented Feb 11, 2026

Applies persuasion principles from research to make critical AI instructions more enforceable.

Changes

  • Added "NOT NEGOTIABLE" declarations to 5 critical sections
  • Added blocking language ("You CANNOT proceed until")
  • Added "1% chance" thresholds to remove AI judgment
  • Added concrete failure modes showing consequences
  • Strengthened Golden Rules with "No exceptions"

Sections Updated

  1. LeanIX MCP Tools - Schema verification enforcement
  2. Golden Rules - Strengthened authority
  3. When You Don't Know - ASK - Added 1% threshold
  4. View Model Colors - Added failure mode
  5. Using Translations - Added failure mode

References

E.g. see how they're making claude code to use skills https://github.com/obra/superpowers/blob/main/skills/using-superpowers/SKILL.md (claude code is lazy to use known skills by default)

Branch Info

This branch is based on PR #48 (CMP-226: Define business logic).

- Rename section to 'When You Don't Know - ASK (Critical)'
- Emphasize: NEVER make assumptions, users prefer questions
- Add 'Uncertain Business Logic' subsection with examples
- Add checklist items: 'No assumptions' and 'Business logic documented'
- Keep guidance concise and actionable
Applied NOT NEGOTIABLE, blocking language, and failure modes to schema verification, colors, and translations sections.
@internal-ai-adoption-analytics

📊 Survey: Please help us track internal AI adoption. This is not for evaluation, only for metrics. You can update this anytime before merge.
🤖 Which AI developer tool(s), if any, were used to assist with this PR? Please tick one or more boxes below:

  • None
  • Cline
  • Claude Code
  • Cursor
  • GitHub Copilot
  • Other

🧑‍💻 If AI was used, what was it mainly used for?

  • Coding / Implementation
  • Testing (unit, integration, e2e)
  • Documentation
  • Research / Investigation
  • Refactoring / Cleanup
  • Configuration / DevOps
  • Other

Thank you in advance for your feedback.💙🙏🎉
Here's a joke to brighten your day: How do you know if an AI is lying? Check its confusion matrix! 📈

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

1 participant