skills/venhdev/zenforge/ci-pub/Gen Agent Trust Hub

ci-pub

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests repository-local documentation files (doc/runbooks/, docs/, CONTRIBUTING.md) to extract project-specific release invariants. While this creates a surface for indirect prompt injection if the files contain malicious instructions, the risk is mitigated by explicit human authorization gates for all manifest mutations and git commits.
  • Ingestion points: SKILL.md Phase 1 (Discovery) reads markdown files from documentation directories.
  • Boundary markers: None explicitly defined for document parsing.
  • Capability inventory: Modifies pubspec.yaml and CHANGELOG.md, seeds GitHub Action workflows, and executes git commit.
  • Sanitization: None specified for the content extracted from documentation.
  • [COMMAND_EXECUTION]: The skill utilizes several standard command-line tools for development and release workflows.
  • Tooling: Uses git for committing and tagging, dart and flutter SDKs for formatting, analysis, and testing, and grep/sed for manifest parsing.
  • Restricted scope: Operating invariants strictly forbid autonomous publication (dart pub publish is only executed with --dry-run).
  • [EXTERNAL_DOWNLOADS]: The skill performs network requests to the official Dart package registry.
  • Service: Fetches version metadata from pub.dev via curl to compare local and remote version baselines. These requests target a well-known, official service for the Dart ecosystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 01:26 PM