← 返回 Skills 市场
480
总下载
1
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install tricore
功能描述
TriCore Architecture - A deterministic (Code-First) three-layer memory and cognitive framework designed for low-resource servers/Agents. It includes the unde...
安全使用建议
This package is internally consistent: it implements a local memory engine (tools/memctl.py), templates for planning/ReAct/self-evolution, and an installer that enforces use of the engine by modifying POLICY.md and OpenClaw config. Before installing, consider: 1) Back up POLICY.md and MEMORY.md (install will move/rename a large MEMORY.md). 2) Review tools/memctl.py and install.sh yourself (they are local and do not fetch remote code). 3) Be aware install.sh will call 'openclaw config set' to override compaction behavior and inject a Linter policy — this is intrusive and affects other skills; test installation in an isolated workspace or sandbox first. 4) The self-evolution skill uses edit/write permissions (it can modify code); only enable it if you trust the skill and are prepared to review/monitor automated code changes. 5) If you are unsure, run the installer manually in a throwaway environment, inspect the created files in memory/ and policy changes, and verify you can revert them with uninstall.sh.
功能分析
Type: OpenClaw Skill
Name: tricore
Version: 1.0.1
This skill bundle is classified as suspicious due to its extensive use of highly privileged tools and its design to perform broad system modifications and agent self-modification. The `install.sh` script directly alters the agent's core `POLICY.md` and OpenClaw's `compaction.memoryFlush.prompt` configuration. Furthermore, the `cognitive-skills/self-evolution.md` skill explicitly instructs the agent to modify its own code and system configurations using powerful tools like `default_api:exec`, `default_api:edit`, and `default_api:write`. While the stated intent is architectural enforcement and self-improvement, these capabilities introduce a significant attack surface and high potential for unintended consequences or exploitation, even without clear evidence of intentional malicious behavior like data exfiltration or backdoor installation.
能力评估
Purpose & Capability
The package implements a code-first memory manager (tools/memctl.py), cognitive-skill templates, and an installer — all consistent with a 'TriCore' memory/cognition framework. Requiring Python, adding memctl, and providing planning/React/self-evolution templates is proportional to the described purpose. However, the installer also injects system policy rules and overrides OpenClaw's compaction prompt (system-wide configuration), which is a stronger footprint than many memory helper skills would need.
Instruction Scope
SKILL.md and the included files explicitly instruct agents to forbid arbitrary file creation, route all persistence through memctl.py, run memctl.py lint for any automation, and use memory_search for retrieval. Those instructions align with the skill's goal (centralize memory), but they place strong constraints on normal agent behavior and tell the agent to read/migrate any existing MEMORY.md (migrate_legacy). The instructions do not request unexplained credentials or network exfiltration.
Install Mechanism
There is no remote download of code; install.sh copies files from the package into the workspace and runs local Python scripts. No external URLs, shorteners, or remote archives are fetched. That lowers supply-chain risk compared to fetching remote binaries. The script does call 'openclaw config set' (a system CLI) which is expected to configure OpenClaw but will fail if OpenClaw isn't present.
Credentials
The skill does not request environment variables, tokens, or credentials. It does, however, recommend (but not require) optional components like an 'agent-browser' and expects the agent to have exec/write/edit permissions to perform self-evolution — this is documented and consistent with the skill's self-modifying/evolution objectives.
Persistence & Privilege
The installer injects a TriCore block into POLICY.md and overrides OpenClaw's pre-compaction memory flush prompt (openclaw config set). Those are persistent, system-wide changes that affect global behavior and other skills. The skill also provides a self-evolution template that uses default_api:edit/write (ability to modify code). While these are coherent with a 'memory OS' designer's goals, they grant broad, persistent influence over the runtime and tooling and could lock or interfere with other workflows.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install tricore - 安装完成后，直接呼叫该 Skill 的名称或使用
/tricore触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.0.1
Added pre-compaction override and bilingual support
v0.1.2
tricore v0.1.2 Changelog
- Added a “System Compatibility (Compaction Hook)” feature to TriCore architecture, automatically patching the OpenClaw prompt for memory compaction to prevent unauthorized file writing loops.
- Updated SKILL.md documentation to include details about this compaction hook and clarify its role in preventing HTTP 429 burst errors during memory flush.
- install.sh and uninstall.sh scripts updated to implement/handle the new compatibility mechanism.
- No breaking changes; existing usage is unaffected, but environments are now more robust against write-related compaction problems.
v0.1.1
TriCore 0.1.1 Changelog
- Added `uninstall.sh` for easier skill removal.
- Updated documentation for clarity and consistency in English.
- Improved and clarified the setup, operation, and compliance instructions across README and cognitive skill files.
- Enhanced guidance on using and enforcing `memctl.py`-based memory operations.
- Minor content and formatting improvements throughout.
v0.1.0
Initial release of TriCore: a deterministic, code-first memory and cognition architecture for low-resource Agents.
- Implements a strict three-layer memory system with unified management via memctl.py.
- Enforces memory access and updates exclusively through command-line tools and semantic search (memory_search), prohibiting ad-hoc file writing.
- Includes built-in linting to block non-compliant file operations or script automation.
- Provides a streamlined installation process with automatic migration of legacy memory files.
- Bundles core cognitive skill templates (planning, ReAct, self-evolution) optimized for the TriCore system.
- Designed for compatibility with OpenClaw (v2026+), Python 3.6+, and key system utilities.
元数据
常见问题
TriCore 是什么？
TriCore Architecture - A deterministic (Code-First) three-layer memory and cognitive framework designed for low-resource servers/Agents. It includes the unde... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 480 次。
如何安装 TriCore？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install tricore」即可一键安装，无需额外配置。
TriCore 是免费的吗？
是的，TriCore 完全免费（开源免费），可自由下载、安装和使用。
TriCore 支持哪些平台？
TriCore 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 TriCore？
由 bertonhan（@bertonhan）开发并维护，当前版本 v1.0.1。
推荐 Skills