general-agentic-memory
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a workflow for ingesting and processing untrusted external data sources, such as PDF documents, video files, and JSONL agent logs, into a searchable memory system.
- Ingestion points: The
wf.add(input_file="...")method and thegam-addCLI command inSKILL.mdare used to import content from local files into the memory framework. - Capability inventory: The skill uses LLMs (e.g., GPT-4o) to chunk, summarize, and query the ingested data through the
wf.request()method and thegam-requestCLI. - Boundary markers: The documentation does not specify the use of delimiters or system instructions to prevent the LLM from following commands embedded within the processed content.
- Sanitization: There is no evidence of filtering or sanitizing the input data before it is processed by the memory building and querying agents.
Audit Metadata