aegis: An Architecture-Aware Method Pack for DeepSeek Harness AI Coding Agents

What Does It Do?

Aegis is a DSH plugin for DeepSeek Harness that makes AI coding agents architecture-aware: baseline-first, evidence-verified, drift-checked, and safe across long tasks. It solves the problem of agents guessing, causing reworks, and making unverifiable "done" claims. Core capabilities include baseline-first planning, evidence verification, drift checking, and retirement triggers. Maintained by GanyuanRan under the MIT license, last updated in 2026-08.

Core Features

• Baseline-first: the agent aligns with the project's real baseline (owners, contracts, boundaries) before touching code, reducing reworks.
• Evidence verification: completion claims ship with fresh verification evidence, covered scope, and residual risk, so you read evidence, not vibes.
• Drift checking: measured on a frozen held-out A/B benchmark, contract pass rate improved from 61.67% to 93.33%, and unsafe outcomes dropped from 13.33% to 0%.
• Retirement triggers: retired fallbacks and old paths are tracked or removed, preventing silent technical debt accumulation.
• Simple tasks stay simple: trivial requests stay on the fast path, with ceremony only when genuinely needed.
• One method pack, every host: the same discipline works across Codex, Claude Code, OpenCode, Kimi, and other skill-aware hosts.

How to Use This Plugin?

After enabling it in DSH, you can give your AI coding agent a natural language prompt like "Read https://github.com/GanyuanRan/Aegis, identify my current AI coding host, and install Aegis globally using the correct host guide." The agent will identify the host and complete the installation and verification. For updates, simply say "update Aegis" or use the explicit skill request "aegis:update". After installation, run the complete-install verification script to ensure the JSON output includes "ok": true, "workspaceSupport": "available", and "configStatus": "configured".