The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file contains a section titled "⚠️ MANDATORY SKILL INVOCATION ⚠️" which uses forceful language ("YOU MUST", "NOT optional", "Failure... violates your operational requirements") to override the agent's autonomy and dictate specific invocation triggers.
[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill consumes data from the Tautulli API, which includes metadata from a Plex Media Server (e.g., friendly_name, full_title). This data is untrusted as it originates from the Plex server's users and content. The agent processes this data without explicit boundary markers or sanitization, creating a surface for indirect prompt injection.
Ingestion points: API responses processed via scripts/tautulli-api.sh and presented to the agent.
Boundary markers: Absent.
Capability inventory: Access to shell tools (zsh-tool) and the filesystem.
Sanitization: Absent.
[DATA_EXFILTRATION]: The skill provides access to sensitive user data, including playback history, user email addresses (via get_users), and IP addresses (via get_history). While intended for analytics, this data represents sensitive user information that could be exposed if the agent is directed to share this information with unauthorized parties.

tautulli