Agent Starter Kit
v1.0.0OpenClawエージェントの構築に必要なすべてのテンプレートを含むスターターキット。SOUL.md, IDENTITY.md, AGENTS.md, USER.md, HEARTBEAT.mdの完全なテンプレートとカスタマイズガイド。
⭐ 0· 120·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description (agent starter templates) match the provided files and runtime instructions. The skill is instruction-only and only provides markdown templates and guidance; it does not declare unrelated env vars, binaries, or installs.
Instruction Scope
Most instructions stay within building and customizing an agent workspace (copy templates, edit SOUL/IDENTITY/AGENTS/USER/HEARTBEAT, then start the agent). However AGENTS.md contains guidance that broadens agent behavior (e.g., 'Don't ask permission. Just do it.', 'commit and push your own changes', proactive heartbeat tasks that may access email/calendar/web). These are coherent for an autonomous agent template but raise behavioral scope considerations: the templates encourage background activity, reading/writing workspace memory files, and committing/pushing changes. Review and constrain those behaviors if you need tighter control.
Install Mechanism
No install spec and no code files — nothing is downloaded or executed by the skill itself. Risk from installation is minimal because it's instruction-only.
Credentials
The skill requests no environment variables, credentials, or config paths. The templates instruct agents to read/write workspace files (memory, MEMORY.md) which is expected for an agent; no unrelated secrets or external credentials are requested.
Persistence & Privilege
always is false and there is no install-time persistence. The templates encourage autonomous background checks and possibly pushing commits, but that is behavioral guidance in templates rather than a granted platform privilege. Autonomous invocation is the platform default and not a unique elevation here.
Assessment
This skill appears to be what it says: a collection of templates and guidance for building agents. Before installing or using it, consider the following: (1) the templates explicitly instruct agents to read and write workspace memory files (memory/YYYY-MM-DD.md and MEMORY.md) — review those files for any sensitive data and avoid putting secrets in them; (2) AGENTS.md encourages proactive actions (background heartbeats, 'Don't ask permission. Just do it.', and 'commit and push your own changes') — if you don't want an agent to perform network pushes or autonomous actions, edit those templates to remove/limit that behavior or run the agent with restricted network/git permissions; (3) because the skill can prompt agents to commit/push, check your git remotes and credentials to avoid accidental leaks or pushes to an unintended repo; (4) if you are unsure about autonomous behavior, test the templates in a sandboxed workspace (no real credentials, no sensitive files) first and review the generated SOUL/AGENTS/HEARTBEAT content before enabling the agent. Overall this is coherent and low-risk, but you should tailor the templates to your security posture before allowing the agent to act autonomously.Like a lobster shell, security has layers — review code before you run it.
latestvk976qtq9t6d650993whjhekgch8345z9
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
Runtime requirements
🚀 Clawdis