Add Agnet input output logic narrative

Add a no-code Chinese explanation of the current Agnet framework logic and implementation process using only Agnet inputs, outputs, handoffs, and convergence behavior.

Constraint: The user asked for a framework logic explanation without code, based on Agnet input and output.

Rejected: Expanding the generated model I/O report again | a standalone narrative keeps the explanation readable and avoids another raw evidence dump.

Confidence: high

Scope-risk: narrow

Directive: Keep this document prose-only; do not add source snippets, command blocks, or inline code markers.

Tested: no backtick/code-marker scan on docs/AGNET_FRAMEWORK_INPUT_OUTPUT_LOGIC.zh-CN.md; .venv/bin/python -B -m unittest discover -s tests; .venv/bin/python -B -m py_compile swarm_minimal/*.py examples/*.py tests/*.py; git diff --check; docs secret pattern scan.

Not-tested: Live S07 was not rerun because this change adds explanatory documentation only.

Co-authored-by: OmX <omx@oh-my-codex.dev>

README.md

@@ -25,6 +25,7 @@
 - `docs/AGENT_SWARM_QUALITY_STANDARD.zh-CN.md`：Agent / 蜂群 Agent 质量标准、测试标准来源、S07 外部 GitHub 代码任务分配和接手机制。
 - `docs/ACADEMIC_STANDARD_TEST_REPORT.zh-CN.md`：S01-S08 测试场景、测试过程、测试结果、算法清单和马尔可夫过程判定。
 - `docs/MODEL_AGNET_IO_REPORT.zh-CN.md`：每个模型 / Agnet 的任务、输入、输出、评分和交接过程。
+- `docs/AGNET_FRAMEWORK_INPUT_OUTPUT_LOGIC.zh-CN.md`：不带代码的框架逻辑说明，只从 Agnet 输入、输出、交接和收敛解释实现过程。
 - `docs/INDUSTRY_STANDARD_AGNET_TEST_PLAN.md`：行业参考框架到本项目标准的映射。
 
 目标是先把蜂群的四个共享资源跑通：