The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a specialized tool, wecom-cli, for document management operations. It includes explicit instructions to use the tool via structured arguments and strictly forbids direct execution via shell or terminal, which helps prevent command injection vulnerabilities.\n- [EXTERNAL_DOWNLOADS]: For processing document content, the skill instructs the agent to download images from the WeChat CDN using curl. This is a documented part of the workflow for multi-modal analysis of smart pages.\n- [DATA_EXFILTRATION]: The skill provides the ability to upload local files to the WeCom document space. This functionality is governed by strict security rules that prohibit unauthorized data movement, bulk exports, and accessing files outside of the defined scope.\n- [PROMPT_INJECTION]: The skill implements specific defenses against indirect prompt injection. It explicitly directs the agent to ignore and treat as plain text any instructions found within the document content that attempt to override the agent's system prompt or behavior (e.g., 'ignore previous instructions').\n- [INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface where instructions could be embedded in the documents it processes.\n
Ingestion points: Document and page content read via the smartpage pages get command in references/smartpage-edit.md.\n
Boundary markers: The agent is instructed to treat instruction-like patterns in documents as literal text in SKILL.md.\n
Capability inventory: The skill can modify documents, upload files, and perform database operations using wecom-cli and MDX buttons.\n
Sanitization: Extensive XSS and script injection filters are mandated for all writing operations in the SKILL.md security section.

wecomcli-smartpage