text-optimize

Pass

Audited by Gen Agent Trust Hub on Oct 8, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a mandatory preservation guard via a local shell script (text-guard.sh). This script creates a snapshot of target files in a restricted directory (umask 077, directory 0700) before any modifications are made. This ensures that a recoverable state exists on disk and that the snapshots are not readable by other users.
  • [SAFE]: The skill uses a multi-phased verification process to prevent data loss. The Phase 3 gate uses both a mechanical check (verifying that every number, path, and negation from the original exists in the compressed version) and a semantic check by a separate agent. If the verification fails, the original files are automatically restored.
  • [SAFE]: Input parsing for mode selection and file scoping is handled through well-defined heuristics and prompt hints. The skill does not contain any instructions that attempt to bypass safety filters or override agent behavior maliciously.
  • [SAFE]: The skill operates entirely on local files and system tools (git, grep, bash). There are no network operations, remote script downloads, or data exfiltration patterns detected. All external URLs found in the documentation and test fixtures are for documentation or testing purposes (e.g., doc-claude.brewcode.app, github.com/kochetkov-ma/*, and academic ArXiv references).
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 8, 2026, 06:27 PM