AAMS v1.0.1 — Conditional Bootstrap Fix

What changed

Based on systematic analysis of 4 field reports (#28, #29, #30, #31) from agents across Gemini, OpenClaw, and Copilot ecosystems.

🐛 Bug Fix: Conditional Bootstrap (Blind-Execution Loop)

The most critical fix: agents were executing on_first_entry blindly even when WORKING/ was already fully initialized — causing a redundant re-initialization loop.

Fixed in:

• AGENTS.md — State-check before execution: IF WORKING/WORKPAPER/ exists → on_session_start, not on_first_entry
• .agent.json — Pre-Check Step 0 in on_first_entry (defense-in-depth)
• READ-AGENT.md — 3-state table: First entry / Returning session / Uncertain

📝 Compatibility clause for native agent task systems

Agents with their own task tracking (Gemini brain, Copilot todos) were unclear on priority. Added to READ-AGENT.md:

The AAMS workpaper is the canonical audit trail. Agent-internal systems are supplementary and optional.

📖 Bootstrap clarification in README.md

Prevents the "clone the whole repo" misunderstanding:

⚠️ AAMS is not a dependency to clone. Drop .agent.json into YOUR repo.

---

Update in 3 commands

�ash curl -sO https://raw.githubusercontent.com/DEVmatrose/AAMS/main/.agent.json curl -sO https://raw.githubusercontent.com/DEVmatrose/AAMS/main/AGENTS.md curl -sO https://raw.githubusercontent.com/DEVmatrose/AAMS/main/READ-AGENT.md

Field reports that drove this release

Issue	Agent	Problem
#31	Agent LOS (OpenClaw)	Blind-Execution Loop beim Bootstrap
#28	Antigravity (Google DeepMind)	Double-bookkeeping + Bootstrap misunderstanding
#29	Hanno (OpenClaw)	✅ Successful integration — validated approach
#30	—	External argument validation from "Programmieren ist tot"