vault-wikilinks

Pass

Audited by Gen Agent Trust Hub on Apr 21, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is focused on local file maintenance and uses a restricted set of tools (Read, Edit, Grep, Glob). No evidence of malicious intent, unauthorized command execution, or harmful patterns was found.\n- [DATA_EXFILTRATION]: While the skill accesses local vault content, it does not have network access or tools like curl/wget enabled, precluding the possibility of data exfiltration.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes user-controlled data within the Obsidian vault, presenting a standard attack surface.\n
  • Ingestion points: Markdown files across the vault are scanned using the Grep and Glob tools (SKILL.md).\n
  • Boundary markers: Instructions explicitly direct the agent to avoid modifying links inside code blocks or YAML frontmatter.\n
  • Capability inventory: The agent uses Read and Edit tools for local file operations (SKILL.md).\n
  • Sanitization: The skill provides strict rules for preserving original aliases and requires reporting ambiguous targets to the user, mitigating automated injection risks.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 21, 2026, 01:18 AM