The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the wacli tool from an external GitHub repository (github.com/steipete/wacli) and a third-party Homebrew tap (steipete/tap/wacli) during setup.
[COMMAND_EXECUTION]: The skill executes multiple CLI commands including wacli auth, wacli sync, wacli messages search, and wacli send, which interact with the local file system and network.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It ingests untrusted data by searching and reading WhatsApp chat history, which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: wacli messages search, wacli chats list, and wacli history backfill retrieve content from external WhatsApp messages.
Boundary markers: The instructions include manual confirmation steps ("Confirm recipient + message before sending"), which mitigate accidental execution but do not fully prevent adversarial injection.
Capability inventory: The skill can send messages (wacli send) and access local files (wacli send file --file /path/agenda.pdf).
Sanitization: No automated sanitization or filtering of the retrieved chat message content is specified before the agent processes it.

wacli