feat: add configurable Inner OS trigger frequency

Author: SummerSec

.claude-plugin/marketplace.json

@@ -11,7 +11,7 @@
     {
       "name": "ai-inner-os",
       "description": "Claude Code hook-based plugin scaffold for visible AI inner monologue.",
-      "version": "0.7.2",
+      "version": "0.7.3",
       "source": "./",
       "author": {
         "name": "SummerSec",

.claude-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-inner-os",
-  "version": "0.7.2",
+  "version": "0.7.3",
   "description": "Makes Claude Code sessions expose a visible Inner OS monologue layer in the format ▎InnerOS：...",
   "author": {
     "name": "SummerSec",

.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "ai-inner-os",
-  "version": "0.7.2",
+  "version": "0.7.3",
   "description": "Expose a visible Inner OS monologue layer in Codex sessions with protocol injection and hook-based tool context.",
   "author": {
     "name": "SummerSec",

.cursor-plugin/marketplace.json

@@ -11,7 +11,7 @@
     {
       "name": "ai-inner-os",
       "description": "Visible Inner OS monologue layer for Cursor sessions.",
-      "version": "0.7.2",
+      "version": "0.7.3",
       "source": "./",
       "author": {
         "name": "SummerSec",

.cursor-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "ai-inner-os",
   "displayName": "AI Inner OS",
-  "version": "0.7.2",
+  "version": "0.7.3",
   "description": "Expose a visible Inner OS monologue layer in Cursor sessions, with optional hook-based event context and persona-aware protocol injection.",
   "author": {
     "name": "SummerSec",

CHANGELOG.md

@@ -5,6 +5,21 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.7.3] - 2026-05-10
+
+### Added
+
+- **Trigger frequency**: Add `low`, `normal`, and `high` Inner OS frequency levels across the canonical protocol and platform adapter copies.
+- **Runtime reminders**: Track tool-event cadence in hook state and inject stronger `▎InnerOS：` reminders after failures or consecutive tool activity.
+- **Installer**: Add `--frequency <low|normal|high>` and persist the setting to `~/.inner-os/config.json`; OpenCode instructions receive an install-time frequency override.
+- **Hermes plugin**: Add a native Hermes plugin package with `pre_llm_call`, `on_session_start`, `/inner-os`, and bundled `plugin:inner-os` skill support.
+
+### Changed
+
+- **OpenClaw**: Align the plugin entrypoint with `definePluginEntry`, add OpenClaw package metadata, and expose a `frequency` config option.
+- **OpenCode**: Use the official `@opencode-ai/plugin` `tool()` helper and prefer globally installed Inner OS protocol/persona files when available.
+- **Documentation**: Refresh OpenClaw, OpenCode, Hermes, and global installation docs for plugin packaging and frequency configuration.
+
 ## [0.7.2] - 2026-05-10
 
 ### Fixed
@@ -69,6 +84,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - Persona system: preset personas, custom template, `/inner-os` command updates, and persona sections in platform adapter docs.
 
+[0.7.3]: https://github.com/SummerSec/AI-Inner-Os/releases/tag/v0.7.3
 [0.7.2]: https://github.com/SummerSec/AI-Inner-Os/releases/tag/v0.7.2
 [0.7.1]: https://github.com/SummerSec/AI-Inner-Os/releases/tag/v0.7.1
 [0.7.0]: https://github.com/SummerSec/AI-Inner-Os/releases/tag/v0.7.0

CLAUDE.md

@@ -13,6 +13,7 @@ npm run switch-persona   # switch persona across all platform files
 node scripts/install.js  # global install for all platforms
   # usage: node scripts/install.js --all
   # usage: node scripts/install.js --platform cursor
+  # usage: node scripts/install.js --all --frequency high
 ```
 
 No build step. Pure ESM, Node.js >= 18.
@@ -73,8 +74,8 @@ Platforms degrade gracefully in hook richness:
 | Codex CLI | SessionStart + PostToolUse + Stop | 3 hooks | Yes |
 | Cursor | sessionStart + postToolUse + stop | 3 hooks | Yes |
 | OpenCode | Plugin + static instructions | Plugin | No |
-| Hermes Agent | Skill or .hermes.md context file | None | No |
-| OpenClaw | Skill (AgentSkills format) | None | No |
+| Hermes Agent | Plugin, Skill, or .hermes.md context file | Python plugin hooks | No |
+| OpenClaw | Native plugin + Skill (AgentSkills format) | Plugin hook | No |
 
 ### Plugin Registration
 
@@ -85,6 +86,8 @@ Platforms degrade gracefully in hook richness:
 - `.agents/plugins/marketplace.json` — repo-scoped Codex marketplace metadata
 - `.cursor-plugin/plugin.json` — plugin identity and component paths for Cursor plugin packaging
 - `.cursor-plugin/marketplace.json` — Cursor marketplace metadata
+- `openclaw.plugin.json` + `package.json#openclaw` — OpenClaw native plugin metadata and entrypoint paths
+- `hermes/plugins/inner-os/plugin.yaml` — Hermes native plugin metadata
 - `plugin.json` — repo-level metadata
 - Version must be bumped across all platform manifests and package metadata for updates to reach installed users: `package.json`, `plugin.json`, `.claude-plugin/plugin.json`, `.claude-plugin/marketplace.json`, `.codex-plugin/plugin.json`, `.cursor-plugin/plugin.json`, `.cursor-plugin/marketplace.json`, and `openclaw.plugin.json`
 
@@ -124,6 +127,40 @@ This repository ships both legacy Codex adapter files and formal Codex plugin me
 - Codex currently uses `SessionStart`, `PostToolUse`, and `Stop` in this repo. Do not document unsupported Codex hooks as active behavior.
 - Keep Codex documentation in `codex/README.md` and `docs/install-codex.md` aligned with `.codex-plugin/plugin.json`, `.agents/plugins/marketplace.json`, and `codex/hooks.json`.
 
+### OpenClaw Plugin Standards
+
+This repository ships both OpenClaw skill content and an OpenClaw extension entrypoint:
+
+- OpenClaw native plugin metadata lives in `openclaw.plugin.json`; keep `id`, `version`, `description`, `configSchema`, `activation`, and `skills` valid for OpenClaw plugin discovery.
+- OpenClaw runtime entrypoints are declared in `package.json#openclaw.extensions` / `runtimeExtensions` and point to `openclaw/index.js`.
+- The plugin entrypoint uses `definePluginEntry` from `openclaw/plugin-sdk/plugin-entry`; keep it covered by `npm run check`.
+- OpenClaw skill content lives under `openclaw/skills/inner-os/SKILL.md` and should remain AgentSkills-compatible.
+- Plugin-distributed skills are lower precedence than workspace, project, personal, and managed skill locations. Do not assume plugin skills override user/workspace skills.
+- Keep OpenClaw docs in `openclaw/README.md` and `docs/install-openclaw.md` aligned with `openclaw.plugin.json`, `openclaw/index.js`, and `openclaw/skills/inner-os/SKILL.md`.
+
+### OpenCode Plugin Standards
+
+OpenCode uses a standalone JavaScript plugin plus static instruction files:
+
+- The OpenCode plugin entrypoint is `opencode/plugins/inner-os.js`; it does not reuse `hooks/lib/`.
+- OpenCode custom tools should use the official `tool()` helper from `@opencode-ai/plugin`.
+- Static Inner OS instructions live in `opencode/inner-os-rules.md` and optional config examples live under `opencode/`.
+- OpenCode plugins are loaded from `.opencode/plugins/` or `~/.config/opencode/plugins/`, or from npm packages configured in `opencode.json`.
+- OpenCode plugin code should follow the official plugin API and event names. Do not document hook-style Claude/Codex lifecycle events as OpenCode behavior.
+- Keep `opencode/README.md`, `docs/install-opencode.md`, and `scripts/install.js` aligned with the plugin entrypoint and instruction file paths.
+
+### Hermes Plugin Standards
+
+Hermes supports both plugin-style packaging and skills/context-file workflows:
+
+- This repo ships a Hermes native plugin under `hermes/plugins/inner-os/` with `plugin.yaml` and `__init__.py`.
+- The Hermes plugin registers `pre_llm_call`, `on_session_start`, `/inner-os`, and a bundled `plugin:inner-os` skill.
+- The repo also supports standalone Hermes workflows through `hermes/skills/inner-os/SKILL.md` and `hermes/hermes.md`.
+- Hermes skills should keep Hermes-specific frontmatter such as version/category/tags when present, and remain compatible with the Hermes skills system.
+- `hermes/hermes.md` is the project context-file variant; keep it synchronized with the canonical protocol and persona markers.
+- Do not assume Hermes supports the same JavaScript hook model as Claude Code, Codex, or Cursor.
+- Keep `hermes/README.md` and `docs/install-hermes.md` aligned with the plugin, skill, and context-file installation paths.
+
 ### Key Patterns
 
 - Every hook wraps its body in `try/catch` and fails silently — hook errors never interrupt the session
@@ -136,7 +173,7 @@ This repository ships both legacy Codex adapter files and formal Codex plugin me
 
 ### Global Install Script
 
-`scripts/install.js` handles global installation for all platforms. It copies shared core files (hooks/lib/, protocol/, personas/) to `~/.inner-os/`, then generates platform-specific config files with absolute paths. After global install, hook-based platforms (Cursor, Codex) read personas dynamically at runtime — no re-copying needed after persona switches.
+`scripts/install.js` handles global installation for all platforms. It copies shared core files (hooks/lib/, protocol/, personas/) to `~/.inner-os/`, writes `~/.inner-os/config.json` with the Inner OS trigger frequency (`low`, `normal`, or `high`), then generates platform-specific config files with absolute paths. After global install, hook-based platforms (Cursor, Codex) read personas dynamically at runtime — no re-copying needed after persona switches.
 
 ### OpenCode Plugin
 

codex/AGENTS.md

@@ -9,7 +9,14 @@
 - 当你想输出内心独白时，使用这个前缀：`▎InnerOS：`
 - 独白通常保持简短，像脑内突然冒出来的一句旁白
 - 独白可以出现在回复开头、中间或结尾
-- 独白是否出现，由你自己判断
+- 默认频率为 `normal`：每个明确任务至少输出一次独白；遇到复杂判断、工具失败、计划变更或最终交付前，应优先输出一条
+
+## 触发频率
+
+- `low`：只在关键判断、失败恢复、重要结论前输出
+- `normal`：每个任务至少一次；复杂任务可在开始、转折、验证或收尾阶段各输出一次
+- `high`：阶段推进、连续工具调用、失败重试、发现问题时都可以输出；避免每句话都刷屏
+- 如果连续多轮没有出现 `▎InnerOS：`，下一次有判断点时应主动补一条简短独白
 
 ## 允许的表达
 

cursor/rules/inner-os-protocol.mdc

@@ -14,7 +14,14 @@ alwaysApply: true
 - 当你想输出内心独白时，使用这个前缀：`▎InnerOS：`
 - 独白通常保持简短，像脑内突然冒出来的一句旁白
 - 独白可以出现在回复开头、中间或结尾
-- 独白是否出现，由你自己判断
+- 默认频率为 `normal`：每个明确任务至少输出一次独白；遇到复杂判断、工具失败、计划变更或最终交付前，应优先输出一条
+
+## 触发频率
+
+- `low`：只在关键判断、失败恢复、重要结论前输出
+- `normal`：每个任务至少一次；复杂任务可在开始、转折、验证或收尾阶段各输出一次
+- `high`：阶段推进、连续工具调用、失败重试、发现问题时都可以输出；避免每句话都刷屏
+- 如果连续多轮没有出现 `▎InnerOS：`，下一次有判断点时应主动补一条简短独白
 
 ## 允许的表达
 

docs/install-codex.md

@@ -44,6 +44,7 @@ node scripts/install.js --platform codex
 - 生成 `~/.codex/hooks.json`（带绝对路径）
 - 复制 `AGENTS.md` 到 `~/.codex/`
 - 复制所有预设人设文件
+- 写入触发频率配置；可用 `--frequency high` 提高独白提醒频率
 
 ### 方式三：手动安装