The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from multiple ingestion points including memory/MEMORY.md, memory/user-profile.md, memory/active-threads.md, memory/pending-tasks.md, memory/learned-today.md, and memory/contacts.md.
Ingestion points: Reads all files within the memory/ directory.
Boundary markers: The instructions do not define boundary markers or delimiters to isolate content during consolidation or archiving.
Capability inventory: The skill has Read, Write, Edit, Glob, and Grep capabilities, allowing it to modify the file system based on processed data.
Sanitization: There is no evidence of sanitization or filtering of the content being moved between files or summarized in the index.
The lack of network access or shell execution limits the severity, but malicious instructions stored in memory could influence how the agent organizes or interprets user data.

memory-manager