abd-context-to-memory
Pass
Audited by Gen Agent Trust Hub on Mar 11, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes subprocess.run within index_memory.py and link_workspace_source.py to orchestrate internal Python scripts and execute the Windows mklink utility for folder junction creation. These operations are essential for the skill's document management functionality.
- [EXTERNAL_DOWNLOADS]: The skill requires several third-party libraries including markitdown (from Microsoft), openai, and faiss-cpu. These are well-known packages from trusted organizations used for the skill's primary document processing and indexing tasks.
- [PROMPT_INJECTION]: This skill implements a Retrieval-Augmented Generation (RAG) system, which introduces a surface for indirect prompt injection.
- Ingestion points: Untrusted document data enters the system through convert_to_markdown.py, which processes files in formats such as PDF, DOCX, and PPTX.
- Boundary markers: Absent; the skill does not wrap retrieved content in delimiters or include warnings for the agent to ignore embedded instructions.
- Capability inventory: The skill performs network operations to OpenAI's API in embed_and_index.py and search_memory.py, and executes system commands for file management in link_workspace_source.py and index_memory.py.
- Sanitization: Absent; document content is converted to markdown and indexed without explicit sanitization or filtering of potential instructions.
Audit Metadata