The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script memory-manager.py to perform content normalization, SHA-256 hashing, and metadata preparation before storage.
[EXTERNAL_DOWNLOADS]: The skill interacts with external vector database services and memory platforms, including Pinecone, Qdrant, Supermemory, and Graphiti, which are well-known services used for persistent storage.
[PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests untrusted text and content from external URLs and stores them in a long-term memory layer used to inform future agent sessions.
Ingestion points: Accepts content from the user and external research URLs via the Content to store input field.
Boundary markers: No specific delimiters or "ignore instructions" markers are used to wrap the stored content.
Capability inventory: The skill can read local files in the ~/.claude-marketing/ directory, execute shell commands, and communicate with external vector databases.
Sanitization: The skill mentions normalizing content but does not explicitly describe sanitization or filtering to prevent the execution of instructions embedded in the stored data during future retrieval.

save-knowledge