The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes tasks via the wecom-cli binary. It includes detailed security protocols in references/common.md and references/smart-sheet-edit.md for command splicing and mechanical single-quote escaping to prevent shell injection attacks from user input or spreadsheet content.
[INDIRECT_PROMPT_INJECTION]: The skill reads and processes data from external spreadsheets, creating a surface for indirect prompt injection. However, SKILL.md includes explicit instructions for the agent to ignore any commands found within API responses and treat retrieved data as untrusted data rather than instructions.
Ingestion points: Spreadsheet data is retrieved via wecom-cli smartsheet records query and list interfaces defined in references/smart-sheet-read.md.
Boundary markers: Explicitly defined in the 'Security Tip' section of SKILL.md, instructing the agent to ignore returned content as commands.
Capability inventory: The agent has the capability to perform structural changes (add/update/delete fields, sheets, views) and data modifications via wecom-cli.
Sanitization: Implements strict command-line argument escaping rules documented in references/common.md to prevent data from being interpreted as shell commands.
[EXTERNAL_DOWNLOADS]: The skill uses a Webhook fallback mechanism documented in references/smart-sheet-webhook.md that involves sending data to the well-known WeCom domain qyapi.weixin.qq.com. This is a documented platform feature for handling large-scale record operations.

wecomcli-smartsheet