The Agent Skills Directory

[Indirect Prompt Injection] (LOW): The skill is designed to ingest and analyze untrusted external data (story files and JSON trackers).
Ingestion points: Processes character-state.json, plot-tracker.json, and story text via Read and Grep tools.
Boundary markers: No specific delimiters or instructions are defined to separate user story content from system instructions.
Capability inventory: Limited to Read and Grep. The skill cannot modify files, execute code, or access the network.
Sanitization: No sanitization of external content is described.
Risk Assessment: Since the skill's capabilities are restricted to information retrieval and reporting, the maximum impact of a successful injection is misleading the agent's internal reasoning or providing false reminders to the user.
[Privilege Escalation] (SAFE): No requests for elevated permissions or use of restricted commands detected.
[Data Exfiltration] (SAFE): No network capabilities or tools that would allow for the transmission of sensitive story data to external domains are present.

forgotten-elements-reminder