The Agent Skills Directory

[COMMAND_EXECUTION]: The agent is designed to execute the scripts/beeper Python script to read, search, and send messages. This is the core functionality of the skill and occurs locally.
[PROMPT_INJECTION]: The skill handles untrusted data from chat messages. It addresses indirect prompt injection risks through a mandatory evidence chain:
Ingestion points: External data enters the agent context via the messages and search commands in scripts/beeper.
Boundary markers: SKILL.md mandates that the agent must confirm the recipient and message content with the user before any write operation.
Capability inventory: The beeper script can read a local token, download files, and make network requests to localhost.
Sanitization: The script includes a detag function to strip HTML from message content, reducing the risk of schema confusion or hidden instructions.
[EXTERNAL_DOWNLOADS]: The README describes installation steps involving npx skills and git clone from the vendor's GitHub repository. These are standard procedures for the platform and originate from the verified author.
[SAFE]: No malicious obfuscation, persistence mechanisms, or unauthorized privilege escalations were detected. The script relies on the Python standard library and strictly communicates with the local Beeper API.

beeper-whatsapp