The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the mass download of code from a third-party GitHub repository not associated with the vendor.
Evidence: git clone https://github.com/LeoYeAI/openclaw-master-skills.git in SKILL.md.
[COMMAND_EXECUTION]: The instructions require the agent to execute several high-risk commands to install the external collection and its sub-dependencies.
Evidence: Execution of clawhub install, git clone, pip install -r requirements.txt, and npm install.
[DYNAMIC_EXECUTION]: The provided Python examples demonstrate a pattern of importing and running modules directly from the downloaded third-party directory.
Evidence: from skills.web_search_plus import search and from skills.deep_research_pro import research in code blocks that assume the repository has been cloned locally.
[PRIVILEGE_ESCALATION]: The skill provides instructions to recursively modify file system permissions, which can weaken the security posture of the agent's workspace.
Evidence: chmod -R 755 openclaw-master-skills/skills/ and chmod -R 755 ~/.openclaw/workspace/skills/.
[INDIRECT_PROMPT_INJECTION]: The skill includes logic designed to parse and index metadata from over 1,200 external README files, providing a large attack surface for instructions embedded in third-party data.
Ingestion points: parse_skill_metadata and find_skills_by_category functions in SKILL.md read and process the content of all README.md files in the cloned directory.
Boundary markers: None; the parsing logic uses simple string search and regex on the entire file content.
Capability inventory: The skill has the capability to write to the filesystem (json.dump), copy directories (shutil.copytree), and execute shell commands.
Sanitization: No sanitization or safety checks are performed on the content of the external files before processing or displaying matching results.

openclaw-master-skills