← Back to Skills Marketplace
347
Downloads
0
Stars
2
Active Installs
1
Versions
Install in OpenClaw
/install temp-publish
Description
Skill Cortex is the system's capability cortex. When lacking ability, it autonomously acquires Skills from ClawHub or GitHub, then releases them after use. E...
Usage Guidance
This skill appears to implement exactly what it claims (a temporary Skill fetch-and-run 'cortex'), but it gives the agent power to search/install third-party Skills via your clawhub CLI and to persist a local cortex.json. Before installing: 1) Ensure you trust the ClawHub/GitHub sources your agent will pull from and that clawhub is the official client. 2) Ask the author (or inspect published SKILL.md) to clarify the contradictions around auto-switching candidates and what reflex mode actually skips — you should be comfortable that installs and execution plans will be shown and require explicit consent. 3) Backup ~/.openclaw/skill-cortex/cortex.json and review its entity-filtering behavior if you have privacy concerns. 4) Consider limiting use to manual invocation until you verify reflex promotion behavior and candidate-switching semantics.
Capability Analysis
Type: OpenClaw Skill
Name: temp-publish
Version: 0.0.1
The OpenClaw AgentSkills skill bundle is classified as benign. It is a meta-skill designed to manage other skills, inherently requiring high privileges. However, it incorporates strong security and privacy features: mandatory user confirmation for all skill installations and system dependency installations, strict exclusion of write/delete/shell operations from reflex mode, and critically, explicit 'Entity Filtering' instructions in SKILL.md and DESIGN.md to strip personal data (PII, sensitive values) from stored signal words, preventing privacy risks. While prompt injection is a general concern for AI agents, the instructions consistently guide the agent towards secure, privacy-preserving behavior and user consent, rather than malicious actions or data exfiltration.
Capability Assessment
Purpose & Capability
The skill requires the clawhub binary and its instructions revolve around searching, installing, running, and uninstalling Skills from ClawHub/GitHub and maintaining a local cortex.json — all of which are coherent with a 'Skill Cortex' that dynamically acquires and caches Skills.
Instruction Scope
SKILL.md explicitly reads and writes a persistent file (~/.openclaw/skill-cortex/cortex.json) and instructs running 'clawhub search/install/uninstall' and occasional 'which <dependency>' checks — these are expected. However, there are contradictory instructions around candidate switching (Phase 3.1 says auto-switch on failure; a later line says switching requires user consent) and reflex behavior: reflex mode can skip the execution-plan confirmation (though it still shows an install notification). Those contradictions widen the agent's discretion and could lead to unexpected autonomous installs/executions if not strictly enforced.
Install Mechanism
This is instruction-only (no install spec), so nothing is written by this skill itself. It relies on the external 'clawhub' CLI to fetch and install other Skills; that dependency is appropriate for the stated functionality but means the actual code fetched and executed comes from third-party sources (ClawHub/GitHub) and is out of this skill's control.
Credentials
No environment variables or unrelated credentials are requested. The design intentionally records only variable NAMES (e.g., TODOIST_API_KEY) and not values in prefrontal 'env_ready' lessons, and the entity-filtering rules aim to avoid storing concrete personal identifiers — this aligns with the cortex's stated purpose.
Persistence & Privilege
The skill persists a local cortex.json and updates it after each invocation (weights, histories, lessons). That is within scope, but reflex mode (which grants faster, lower-friction invocation) can skip the execution-plan confirmation step; combined with autonomous invocation being allowed by default, this increases the blast radius if reflex promotion or the cortex file is manipulated. Also note the inconsistent language about auto-switching vs. requiring consent for candidate switching — this ambiguity affects how much autonomy the skill will exercise.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install temp-publish - After installation, invoke the skill by name or use
/temp-publish - Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.0.1
Initial release of Skill Cortex, an adaptive short-term capability manager.
- Autonomously acquires and manages Skills from ClawHub or GitHub to fill capability gaps, executing tasks and reinforcing successful flows as reflexes.
- Introduces a multi-phase pipeline: perception (intent matching or search), validation/authorization (with reflex and standard modes), execution (install, plan, run, recover), learning, and cleanup.
- Enforces strict safety and user confirmation rules for installation, write operations, and reflex path eligibility.
- Continuously restructures its memory and patterns based on results, promoting, pruning, and refining skills for optimal evolution.
- Provides detailed error handling, structured lesson recording, and automated cleanup/retention logic for short-term Skills only.
Metadata
Frequently Asked Questions
What is Temp Publish?
Skill Cortex is the system's capability cortex. When lacking ability, it autonomously acquires Skills from ClawHub or GitHub, then releases them after use. E... It is an AI Agent Skill for Claude Code / OpenClaw, with 347 downloads so far.
How do I install Temp Publish?
Run "/install temp-publish" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Temp Publish free?
Yes, Temp Publish is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Temp Publish support?
Temp Publish is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Temp Publish?
It is built and maintained by Ank Wu (@ankwu001); the current version is v0.0.1.
More Skills