dgr
由 sapenov 提供
Web & Frontend Development简介
Audit-ready decision artifacts for LLM outputs — assumptions, risks,.
描述
该技能的功能
DGR produces a structured JSON artifact for each decision, capturing assumptions, risks, a recommendation, and a consistency check. It runs in three modes (dgr_min, dgr_full, dgr_strict) to match the stakes of the decision. Output is schema-validated and intended for storage in audit logs, tickets, or incident records.
Unlike freeform reasoning, DGR enforces a machine-validated schema so decision records are consistent, searchable, and storable without manual formatting.
何时使用
- Logging vendor selection rationale in a project ticket
- Documenting assumptions before a production deployment
- Recording reasoning behind an incident response call
- Capturing risk analysis for a policy change
- Generating an audit-ready artifact for a budget allocation decision
示例工作流
User asks: Should we migrate our authentication service to OAuth2? (mode: dgr_full)
- 1
Accepts the decision request and mode selection
- 2
Identifies decision context and flags any missing key inputs
- 3
Enumerates explicit assumptions and risks with associated severity
- 4
Generates a recommendation with rationale and sets review_required flag accordingly
- 5
Runs a consistency check and returns a schema-valid JSON artifact
A single JSON artifact containing context, assumptions, risks, recommendation, and consistency_check fields, ready for storage in a ticket or audit log
安装
clawhub install sapenov/dgr运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。