The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's metadata contains installation instructions for an external CLI tool wacli from a third-party repository (github.com/steipete/wacli/cmd/wacli@latest). This tool is not from a recognized vendor.
[COMMAND_EXECUTION]: The skill relies on executing shell commands (wacli auth, wacli sync, wacli messages search, wacli send text, etc.) to interact with the local system and the WhatsApp service.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it processes uncontrolled external data by searching and reading incoming WhatsApp messages (wacli messages search).
Ingestion points: wacli messages search and wacli chats list commands in SKILL.md.
Boundary markers: None provided to distinguish between user instructions and data content.
Capability inventory: The agent has the ability to send text messages and upload local files (wacli send text, wacli send file) based on processed data.
Sanitization: No sanitization or filtering of external message content is mentioned.
[DATA_EXFILTRATION]: The skill provides tools to read sensitive communication history and send files (wacli send file). While intended for messaging, this read-and-send capability could be abused for exfiltration if the agent is tricked by malicious data or instructions.

wacli