The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements a persistent memory system that ingests user-controlled narrative content (character descriptions, dialogue, scene notes) and stores it in a local JSON file. This data is later retrieved and interpolated into the agent's context for character validation and memory queries, presenting a surface for indirect prompt injection if the agent is used to analyze untrusted story files.
Ingestion points: User-provided strings for character arcs, attitudes, and scene descriptions stored in .writer-memory/memory.json.
Boundary markers: Absent. The provided logic in lib/synopsis-builder.ts and lib/character-tracker.ts assembles markdown profiles from stored data without explicit delimiters or warnings to the LLM to ignore instructions embedded in the narrative.
Capability inventory: The skill's capabilities are limited to reading and writing within its own local directory via lib/memory-manager.ts. No network or arbitrary code execution tools are present.
Sanitization: The skill performs no filtering or sanitization of input strings to prevent instructions from being stored in memory.
[COMMAND_EXECUTION]: The library code uses standard Node.js fs modules to manage its storage. It follows best practices for local data management, including atomic writes using temporary files and a rotation-based backup system (maximum 20 backups) implemented in lib/memory-manager.ts. All operations are confined to the .writer-memory directory within the user's project.

writer-memory