
Agentic Engineering
https://github.com/shanraisshan/agentic-engineering- Category
- AI Agents
- Rank
- No. 2037Tools index
Previous survey · No. 2009 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- shanraisshan
- GitHub
- 23 stars
- Date
About
Hands-on exploration of agentic engineering practices across Claude, Codex, Cursor, Gemini, and other coding agents.
What it does
A research notebook and configuration bundle for comparing AI coding assistants. It organizes model and context-window findings, defines specialist research roles, and includes Python hooks that add event logging, context injection, and audible notifications to coding-agent sessions.
Why it's ranked here
The project is useful as a worked example of agent-assisted research and hook configuration, especially because it exposes both the research workflow and its outputs. Its value is narrower as a reusable tool: much of the content is reference material, while the executable pieces mainly provide notifications and logging.
What's good
The comparison report records model limits, pricing, selection behavior, and retirement details in one consistent structure. Four specialized research roles divide coverage by provider. Hook configurations cover many lifecycle events, support personal overrides, validate sound names, and fail without interrupting agent work.
Tradeoffs
This is not a standalone coding agent or a packaged research application. Several tables contain fast-changing provider data and therefore require continued verification. The Codex hook documentation also promises detailed session context, while the supplied implementation emits only a fixed short message. Audio support depends on available platform players.
How to use it well
Use it as a reference structure for teams maintaining coding-agent comparisons or experimenting with lifecycle notifications. Adapt the research roles, approval gate, and local hook overrides to your repository. It does not replace source verification, an agent runtime, or a general automation framework.
Technical notes+
CLAUDE.md defines a four-agent, parallel research workflow that writes timestamped results and requires approval before updating README.md. .claude/agents/readme-updater-agent.md supplies the coordinating role, while reports/context-comparison.md holds the detailed comparison. .codex/hooks.json registers five command hooks backed by .codex/hooks/scripts/hooks.py; that script parses JSON from stdin, supports local-over-default configuration, appends selected fields to a JSON Lines log, and plays WAV or MP3 notifications. Its SessionStart path currently returns the fixed text hooks context: run, which conflicts with the richer branch, status, time, and directory context described in .codex/hooks/HOOKS-README.md. .claude/settings.json registers 27 Claude lifecycle events against .claude/hooks/scripts/hooks.py and enables all project MCP servers.
Observed
- Repository type
- Research, comparison, documentation, and coding-agent hook configuration repository
- Executable language
- Python 3
- Install surface
- Direct Python scripts invoked through JSON project configuration; Python 3 is required
- Interfaces
- Claude Code and Codex CLI lifecycle hooks, Markdown reports, and agent definitions
- Platform support
- Hook audio handling covers Windows, macOS, and Linux
- Research structure
- Four specialized research agents cover Claude Code, Codex CLI, Cursor, and Gemini CLI
- Update control
- The documented workflow requires user approval before changing the main comparison
Read from README.md, .codex/hooks/scripts/hooks.py, .claude/hooks/scripts/hooks.py, CLAUDE.md, .codex/hooks.json, reports/ai-terms.md, .claude/settings.json, .vscode/settings.json, reports/context-comparison.md, .codex/hooks/HOOKS-README.md, .claude/hooks/HOOKS-README.md, .claude/agents/readme-updater-agent.md, research/2026-05-02_14-38-59/result.md.
What it can do
Generate code using Claude AI agent
Natural language prompts or coding requirements → Generated source code
Generate code using Codex AI agent
Code descriptions or programming tasks → Generated code snippets or complete functions
Perform AI-assisted coding with Cursor
Coding context and user instructions → Enhanced code with AI suggestions and completions
Generate code using Gemini AI agent
Programming specifications or natural language descriptions → Generated programming code
Compare coding agent performance
Same coding task across multiple AI agents → Performance analysis and comparison results
Explore agentic engineering best practices
Coding scenarios and agent interactions → Documented best practices and implementation patterns
Intel on Agentic Engineering
- #501 – DHH: Future of Programming, AI, Agentic Engineering, Vibe Coding & Linux
- BC-Bench: Evaluating Agentic Engineering in a Domain-Specific Language for ERP
- CCA Exam's Five Domains for Agentic Engineering
- Three closing recommendations for agentic engineering
- Multiplayer agentic engineering — Arjun Singh, Superconductor
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.