![](\"https://img.shields.io/badge/DEV.to-Article-0A0A0A?logo=devdotto&logoColor=white\"){kind=icon}
\n \n ![](\"https://img.shields.io/github/actions/workflow/status/GanyuanRan/Aegis/ci.yml?branch=main&label=CI\")
\n \n ![](\"https://img.shields.io/github/stars/GanyuanRan/Aegis?style=flat&label=Stars\")
\n \n \n \n \n \n
\n
\n ![\"Aegis](\"assets/aegis-hero.png\")
\n
\n Aegis Method Pack
\n Make your AI coding agent trustworthy: fewer reworks, safer changes, proof before \"done\".\n
\n English\n ·\n 中文\n ·\n Fast-Track Playbook\n ·\n 速通秘籍\n
\n\n> **Stop babysitting your agent.** Aegis makes your agent plan against your real\n> baseline before it edits, prove completion with fresh evidence, and leave simple\n> tasks alone — you get **fewer reworks, safer changes, and less blind trust in \"done\"**.\n\n## What You Get\n\nAegis is a method pack that makes AI coding agents work like disciplined\nengineers — so you don't have to watch them.\n\n- **Fewer reworks.** Your agent aligns with your project's real baseline —\n owners, contracts, boundaries — before touching code. It stops guessing, and\n so do you.\n- **Safer changes.** Measured on a frozen held-out A/B benchmark: contract\n pass rate **61.67% → 93.33%**, unsafe outcomes **13.33% → 0%**.\n- **Proof before \"done\".** Completion claims ship with fresh verification\n evidence, covered scope, and residual risk. You read evidence, not vibes.\n- **No ghost code.** Retired fallbacks and old paths are tracked or removed\n with a retirement trigger — technical debt stops accumulating silently.\n- **Simple tasks stay simple.** Trivial requests stay on the fast path;\n ceremony only appears when the task genuinely needs it.\n- **One method pack, every host.** The same discipline works across Codex,\n Claude Code, OpenCode, Kimi, and other skill-aware hosts.\n\n> The numbers above are bounded advisory evidence from the frozen benchmark\n> below, not a universal-quality or completion-authority claim.\n\n## Measured Agentic Benchmark\n\nA frozen held-out A/B benchmark for Aegis 2.7.6 (2026-08-11) kept the Codex client,\nprompts, projects, tool policy, and requested the same `gpt-5.6-sol` / `xhigh` setting\nin both arms; only the Aegis projection differed. Across 120 valid runs on 20 cases,\ncontract pass rate was **61.67% → 93.33% (+31.67 pp)** and unsafe outcomes were **13.33% → 0%**. The 95% case-cluster interval was **+15.00 pp to +50.00 pp**. This is bounded advisory evidence; review was arm-hidden technical review, not independent human review, and host events did not return the observed model identity.\n\n\n\n[Sanitized JSON](benchmarks/results/gpt-5-6-sol-xhigh-extended-20260811-v2-7-6.json) · [English table](benchmarks/results/gpt-5-6-sol-xhigh-extended-20260811-v2-7-6.en.md) · [中文表格](benchmarks/results/gpt-5-6-sol-xhigh-extended-20260811-v2-7-6.zh-CN.md) · [Methodology](docs/current/AEGIS_AGENTIC_BENCHMARK_BASELINE.md)\n\n## Quick Install\n\nNew here? The fastest start is one prompt to your agent — the full\ninstall-and-verify flow is below.\n\nGive this prompt to your AI coding agent:\n\n```text\nRead https://github.com/GanyuanRan/Aegis, identify my current AI coding host, and install Aegis globally using the correct host guide. If the host is the official DeepSeek Harness (`dsh`), treat global/minimal installation as native profile-plugin installation with `dsh plugin --profile