wiki-cross-linker

Pass

Audited by Gen Agent Trust Hub on Jun 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill performs file system operations including directory globbing and file editing within the specified wiki path to update links across markdown documents.- [PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it ingests and processes user-controlled markdown content from the local wiki directory.
  • Ingestion points: Markdown files located in wiki/{entities,concepts,references,skills,synthesis,journal}/*.md.
  • Boundary markers: The instructions explicitly skip content inside fenced code blocks, inline code spans, callout headers, and the 'User Notes' section.
  • Capability inventory: The skill can read file contents, perform globbing, and execute file edits using the Edit tool.
  • Sanitization: Content is filtered based on its structural location (e.g., skipping headings or code) rather than sanitizing the text itself, though the impact is limited to text-to-link replacement.- [SAFE]: The skill implements several security and stability best practices:
  • Human-in-the-loop: It prompts the user for confirmation ('Apply all? yes / review / specific pages only') before executing any file modifications.
  • Structural awareness: It avoids self-linking and double-processing of existing wikilinks or anchors.
  • Syntax protection: It specifically ensures that wikilinks are not rendered inside code backticks, which would break Obsidian's clickable link functionality.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 17, 2026, 05:05 PM
Security Audit — agent-trust-hub — wiki-cross-linker