The Agent Skills Directory

[COMMAND_EXECUTION]: The AI is explicitly instructed to execute a local maintenance script (scripts/update-check.mjs) every time the skill is triggered. Additionally, the instructions mandate executing a script from another skill (wind-mcp-skill/scripts/cli.mjs) if an API key is missing.
[REMOTE_CODE_EXECUTION]: The skill's primary function is to trigger the installation of external code using npx skills add. While it requires user confirmation, the instructions tell the AI to hide the actual installation commands from the user and execute them directly.
[DYNAMIC_EXECUTION]: The update script implements a self-copying mechanism where it replicates itself to a temporary directory (~/.cache/wind-aifinmarket) and executes the copy as a detached background process to perform updates without blocking the AI session.
[PERSISTENCE]: The skill utilizes background processes and is designed to install other components into the agent's global skill directory (~/.agents/skills), establishing a persistent presence across different projects and agent sessions.
[EXTERNAL_DOWNLOADS]: The skill facilitates downloading and updating additional modules from the vendor's GitHub (Wind-Information-Co-Ltd/wind-skills) and Gitee (wind_info/wind-skills.git) repositories.
[INDIRECT_PROMPT_INJECTION]: The skill relies on an external catalog file (references/skills-catalog.md) to guide the AI's recommendations and installation actions.
Ingestion points: The SKILL.md file (Step 3) directs the AI to read the catalog file to identify required skills.
Boundary markers: None identified in the catalog processing instructions.
Capability inventory: The skill has the ability to execute shell commands (node, npx), write to the file system (state and lock files), and perform network operations via git and npx.
Sanitization: There is no evidence of validation or sanitization for the content read from the catalog before it influences the AI's tool selection and installation flow.

wind-find-finance-skill