wiki-status

Pass

Audited by Gen Agent Trust Hub on Nov 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill invokes local command-line utilities including obsidian-wiki, qmd, and scripts/manifest.py to perform graph analysis, search indexing, and data migration tasks. These tools are used within the scope of the skill's maintenance and reporting functions.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: Accesses sensitive local directories such as ~/.claude and ~/.codex to analyze conversation history as part of the wiki status overview. This data is read locally to compute ingested versus pending changes and is not transmitted to external domains.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes third-party or user-generated Markdown notes and conversation logs to generate insights and summaries, which presents a potential injection surface.\n
  • Ingestion points: Processes all .md files in the Obsidian vault and .jsonl history logs from supported AI agents.\n
  • Boundary markers: The skill applies writing profiles but does not explicitly instruct the agent to ignore directives embedded within the source documents being audited.\n
  • Capability inventory: Reads local file systems and executes specialized maintenance CLI tools for graph and cost reporting.\n
  • Sanitization: Uses standard JSON and Markdown parsers to aggregate metadata for status reports.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 17, 2026, 05:31 PM