graph-colorize

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from the user's vault to generate configuration data.
  • Ingestion points: The skill reads tags and folder names from all markdown files ($VAULT_PATH/**/*.md) within the vault.
  • Boundary markers: The instructions specify quoting for paths but do not provide explicit requirements for sanitizing or escaping characters in tags that might break JSON syntax or Obsidian's search query logic.
  • Capability inventory: The skill possesses file-read and file-write capabilities specifically targeting the .obsidian/graph.json and log.md files.
  • Sanitization: There is a lack of formal sanitization for tag-based queries, which could lead to malformed configuration files if tags contain unexpected characters (e.g., quotes).
  • [COMMAND_EXECUTION]: No shell commands or system-level execution patterns were identified; the skill focuses on local file manipulation.
  • [DATA_EXFILTRATION]: The skill operates entirely within the local vault environment and does not attempt any network requests or external data transfers.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 10:32 AM
Security Audit — agent-trust-hub — graph-colorize