harness
Warn
Audited by Snyk on Aug 21, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (medium risk: 0.30). Harness execution (Session 2) invokes
claude-agent-invoke.sh, which reads the host-authored prompt files but also passes the checkpoint’s generatedcontext.md, checkpoint spec, and git diffs as prompt content into the review/Generator agents, and any outsider-authored free text could be included via user-controlled repository files (e.g., docs/code/commit messages) that are LLM-readable through those artifacts.
Issues (1)
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
Audit Metadata