ClawHub

Ai Company Modularization

v1.0.0

AI Company 模块化流程 Skill — 将单体式 Skill 分解为原子化、可独立测试、可组合的模块单元。包含模块边界识别、接口定义、共享逻辑提取、独立版本控制六步流程。

⭐ 0· 46·1 current·1 all-time
by@johnsmithfan
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description (modularizing Skills) align with what the skill requests and instructs: it operates on Skill files, proposes module extraction, interface definitions, dependency analysis and test artifacts. Declared dependencies on related governance skills are plausible.
✓
Instruction Scope
SKILL.md contains step-by-step, local-file-focused procedures, examples, and templates for splitting a Skill into modules. It references reading/writing files under skills/ (which matches declared file permissions) and does not instruct network calls, reading unrelated system paths, or exfiltrating data.
✓
Install Mechanism
No install spec or code files are present; this is instruction-only so nothing is downloaded or written to disk by an installer.
✓
Credentials
The skill requires no environment variables, credentials, or external config paths. That is proportionate for a design/process guidance skill that manipulates repository files.
ℹ
Persistence & Privilege
always:false and no global installs—good. The skill does request read/write access to skills/ (appropriate for modifying Skill modules). Be aware that with autonomous invocation enabled, the agent could create/modify files under skills/ when the skill runs; this is expected but worth controlling.
Assessment
This skill is an instruction-only guide to break a Skill into modules and only asks to read/write files under the skills/ workspace—behavior that is consistent with its purpose. Before installing: (1) review the SKILL.md yourself to confirm you want the agent to create/modify files in your skills/ directory; (2) back up your repository or run the skill in a sandbox/worktree; (3) limit autonomous runs or require user confirmation if you don't want the agent to make changes without review; (4) verify the referenced dependencies (skill-vetter, ai-company-standardization) are trusted. No network or secret access is requested, and there is no installer, so the main risk is unintended local file changes.

Like a lobster shell, security has layers — review code before you run it.

latestvk972j73tnean7mdgexe2x4kw5184s3qp

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments