kb-compile

Pass

Audited by Gen Agent Trust Hub on Apr 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The agent is instructed to run the local script ../obsidian-notes-karpathy/scripts/scan_compile_delta.py to detect changes in the knowledge base. This is part of the skill's functional logic for managing document deltas.
  • [PROMPT_INJECTION]: The skill processes untrusted content from markdown files and PDFs, creating an indirect prompt injection surface. 1. Ingestion points: raw data folders (raw/human/**, raw/agents/**, raw/*.md). 2. Boundary markers: None specified in the instructions for content processing. 3. Capability inventory: Local script execution and file system writes to the wiki/drafts/ directory. 4. Sanitization: No sanitization or content validation is mentioned.
  • [SAFE]: All file operations and script executions are targeted at local or sibling directories within the user's workspace, matching the skill's primary purpose of knowledge management and indexing.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 8, 2026, 07:53 AM