The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the agent's shell capability to run a Python script named "memory-manager.py" with multiple flags to perform data diffing, metadata preparation, and storage operations.
[REMOTE_CODE_EXECUTION]: The skill relies on an external script "memory-manager.py" that is not included in the provided source files. This makes the execution logic unverifiable and relies on pre-existing or external executables on the host.
[DATA_EXFILTRATION]: The skill accesses sensitive information located in the user's home directory ("~/.claude-marketing/"), including brand profiles and campaign insights, and transmits this data to external vector databases such as Pinecone and Qdrant.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection through the processing of untrusted data from session learnings. Ingestion points: Data is gathered from "insights.json", the "campaigns/" directory, and the session context. Boundary markers: Absent; the skill provides no delimiters or instructions to ignore potential commands embedded in the synced content. Capability inventory: Includes local file reads, shell command execution via "memory-manager.py", and external network storage operations. Sanitization: Absent; there is no indication of content validation or escaping before synchronization.

sync-memory