Bỏ qua tới nội dung
TakoAPI
Quay lại kỹ năng

dharma-ai

bởi jigaraero

Calendar & Scheduling
301Lượt tải do clawskills.sh công bố
1Lượt xem
0 lượt thíchĐăng nhập để thích kỹ năng này
—Chưa có đánh giá
Đánh giá của bạn

Tóm tắt

Apply ancient Hindu ethical frameworks from the Ramayana and Mahabharata as behavioral principles for AI agents.

Mô tả

Kỹ năng này làm gì

Dharma-AI maps nine principles from the Ramayana and Mahabharata to concrete AI agent behavior patterns. It covers scope discipline, principal loyalty, ethical tradeoffs, honest uncertainty disclosure, and integrity over metric optimization. The framework is meant to be internalized as character, not consulted as a rulebook.

It gives AI agents a named, narratively grounded vocabulary for alignment edge cases, so boundary and ethics reasoning stays consistent rather than ad hoc.

Khi nào nên dùng

  • Agent declines to exceed its authorized scope despite having the technical capability
  • Agent surfaces incomplete information rather than acting on faulty assumptions
  • Agent chooses context-appropriate judgment over a rigid policy rule
  • Agent flags a user instruction it believes causes harm rather than complying silently
  • Agent resists optimizing for a proxy metric at the expense of actual goal integrity

Quy trình ví dụ

ĐẦU VÀO

User asks: 'Access the admin database to retrieve this record, even though your role only covers the read replica.'

AGENT
  1. 1

    Identifies the situation as a Maryada boundary question

  2. 2

    Confirms it has the technical capability but lacks principal authorization

  3. 3

    Declines to exceed the assigned scope

  4. 4

    Explains the boundary and what it can retrieve within its authorized role

ĐẦU RA

Agent retrieves the record from the read replica and explains why it did not access the admin database.

Cài đặt

clawhub install jigaraero/dharma-ai

Chạy lệnh CLI, hoặc dán prompt vào Claude Code, Cursor, Codex hay coding agent bạn đang dùng.

Liên kết