harness-engineering

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it reads and processes instructions from untrusted repository files (such as README.md, CONTRIBUTING.md, and .github/instructions/) to create durable agent rules.
  • Ingestion points: The skill reads various repository files including documentation (README.md, CONTRIBUTING.md, docs/), agent-specific files (AGENTS.md, .github/copilot-instructions.md), and configuration manifests (package.json, pyproject.toml, etc.).
  • Boundary markers: The instructions do not specify any explicit delimiters or warnings to ignore embedded instructions when reading these files.
  • Capability inventory: The skill is capable of writing new files to the repository, creating executable scripts (Python/Shell), and modifying CI workflows (.github/workflows/).
  • Sanitization: There is no requirement mentioned for validating or sanitizing the content ingested from the target repository before it is used to generate or update instructions.
  • [EXTERNAL_DOWNLOADS]: The skill references an external GitHub repository (https://github.com/baskduf/harness-starter-kit) as a reference implementation for the harness-engineering workflow.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 10:12 AM
Security Audit — agent-trust-hub — harness-engineering