wiki-rebuild

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructions specify the use of shell commands to manage a search index using the qmd tool. These commands are executed during rebuilding and restoration phases.
  • Evidence: Use of ${QMD_CLI:-qmd} update, ${QMD_CLI:-qmd} embed, and ${QMD_CLI:-qmd} ls "$QMD_WIKI_COLLECTION" in the QMD Refresh section.
  • Evidence: Verification step involving ${QMD_CLI:-qmd} get "qmd://$QMD_WIKI_COLLECTION/<restored-page>.md" -l 5.
  • [INDIRECT_PROMPT_INJECTION]: The skill functions by ingesting and restoring previously archived markdown content, which represents a potential attack surface.
  • Ingestion points: Content is read from the _archives/ directory during Mode 3 (Restore from Archive) operations described in SKILL.md.
  • Boundary markers: Absent. The skill does not implement delimiters or specific warnings to ignore instructions embedded within the restored wiki pages.
  • Capability inventory: The skill has the ability to delete live wiki content, copy files between directories, and execute shell commands via the qmd CLI.
  • Sanitization: Absent. The skill restores archived data verbatim to the live vault without validation or filtering of the file contents.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 03:12 PM
Security Audit — agent-trust-hub — wiki-rebuild