Ahmed Hemedan
AI engineer building agent skills for Claude Code, Codex and other coding agents. I work on the hard parts of agentic engineering: context engineering, spec-driven development, verification gates that cannot fake a pass, and long-running agent workflows that do not rot. Every skill runs in my own daily work, with real failure modes turned into rules. For engineers who want agents that ship correct code, not just fast code.
Skills
Context Rot Fix
Fix context rot in long AI projects. Use when a session is getting long, slow, forgetful or contradictory, or when the user says resume, continue where we left off, catch me up, checkpoint, save state, wrap up or hand over. Keeps three state files (CONTEXT.md, STATE_NOW.md, LEDGER.md) so a fresh chat starts clean and resumes in one read.
Spec Driven Agent Testing
Get quality, not just speed, from a coding agent. Use when briefing Claude Code or another agent on a feature, writing tests or CI gates, reviewing agent-written code, hardening a vibe-coded prototype, or whenever an agent reports that all tests pass. Define correct behaviour first, make rules structural, prove every gate can fail.