add harness framework

Codex/plugins/agents-md-management v2/1.0.0/README.md

@@ -0,0 +1,58 @@
+# AGENTS.md Management
+
+A local Codex adaptation of Anthropic's
+[`claude-md-management`](https://github.com/anthropics/claude-plugins-official/tree/main/plugins/claude-md-management)
+plugin.
+
+This derivative is not an Anthropic-published Codex plugin. It preserves the
+original Apache-2.0 license and adapts the two workflows exclusively for
+Codex's `AGENTS.md` and `AGENTS.override.md` instruction system.
+
+## Skills
+
+### `auditing-agents-md`
+
+Audits repository guidance against current code, commands, paths, scope,
+override precedence, duplication, and the instruction byte budget. It produces
+a 100-point quality report and proposed diffs before asking for edit approval.
+
+Example:
+
+```text
+Audit the AGENTS.md guidance in this repository.
+```
+
+### `revising-agents-md`
+
+Reviews the current session for verified, durable learnings and proposes the
+narrowest appropriate `AGENTS.md` destination. It excludes sensitive,
+unverified, generic, transient, and duplicate content.
+
+This skill is explicit-only. Invoke it from the skill picker or as
+`$revising-agents-md`.
+
+Example:
+
+```text
+Use $revising-agents-md to propose durable AGENTS.md updates from this session.
+```
+
+## Safety
+
+Both skills:
+
+- show evidence and exact diffs before editing;
+- require approval for the proposed destinations and content;
+- edit only approved instruction files;
+- verify the resulting instruction chain after a change.
+
+They do not operate on Claude instruction files and include no hooks, MCP
+servers, apps, network integrations, or credential handling.
+
+## Attribution
+
+Original plugin copyright and authorship belong to Anthropic. The original
+plugin was authored by Isabella He and published in Anthropic's official Claude
+plugin repository. This Codex-specific adaptation was generated locally from
+upstream commit `a28630018d1f8ed486d5a88169ac1f2fb8df89ba`.
+