The Agent Skills Directory

[PROMPT_INJECTION]: The SKILL.md file uses directive language ("⚠️ MANDATORY SKILL INVOCATION ⚠️") and claims that failure to invoke the skill violates operational requirements, which is a technique intended to bypass the agent's standard tool-use reasoning.
[COMMAND_EXECUTION]: The scripts/plex-api.sh helper script executes shell commands via curl based on user-supplied parameters. While search terms are URL-encoded, other inputs like library IDs and metadata keys are interpolated into the URL string without strict validation.
[PROMPT_INJECTION]: Documentation in the references/ directory includes examples for destructive API operations like deleting media items and modifying server preferences, which contradicts the skill's stated read-only purpose and creates a risk of unintended data loss.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted media metadata (such as titles and summaries) from the Plex API without boundary markers or sanitization. Ingestion points: metadata returned by scripts/plex-api.sh from the configured Plex server; Boundary markers: Absent; Capability inventory: Shell and network access via curl; Sanitization: No sanitization is applied to the data retrieved from the API.

plex