The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses python3 to execute local utility scripts (memory_store.py and vector_search.py) for processing memory searches and retrieving specific records. These scripts are part of the skill's own package and perform legitimate search operations.
[DATA_EXFILTRATION]: Accesses a local database at ~/mnt/.claude/.cowork-mem/memory.db. This operation is limited to reading the skill's own persistent memory records to provide historical context. No evidence of unauthorized access to sensitive system credentials or external data transmission was found.
[PROMPT_INJECTION]: The skill processes data from past interactions stored in its memory database, which presents an indirect prompt injection surface. Because the retrieval of historical context is the primary intended purpose of the skill and no dangerous downstream capabilities (such as arbitrary code execution or network sends) are triggered by the search results, this is a standard operational risk rather than a malicious finding.
Ingestion points: ~/mnt/.claude/.cowork-mem/memory.db accessed via search scripts in SKILL.md.
Boundary markers: None specified in the search query templates.
Capability inventory: Local script execution for data retrieval; agent synthesizes retrieved context to inform responses.
Sanitization: No explicit sanitization of retrieved memory content is defined in the instruction markdown.

mem-search