ship

Fail

Audited by Snyk on Nov 19, 2026

Risk Level: CRITICAL
Full Analysis

CRITICAL E004: Prompt injection detected in skill instructions.

  • Potential prompt injection detected (medium risk: 0.30). The visible content is a disclosed, gated release workflow with no malicious intent, but its merge instructions tell the agent to invoke merge.sh by a relative path because auto mode refused the absolute form, which works around a harness refusal on a merge action.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

  • Third-party content exposure detected (medium risk: 0.30). The required runtime workflow reads outsider-writable GitHub and Azure DevOps issues and PR review threads and comments (via read-issue, poll-pr and the unattended select) without the agent first choosing a specific item, so an outsider can post into a feed the workflow consumes.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

  • Potentially malicious external URL detected (medium risk: 0.30). Setup installs the composed skills from GitHub repositories pinned to commit SHAs, so their content is operational; the mattpocock/skills source is an individual-owned repository (unofficial) whose immutable SHA pin keeps it from being weak, giving a reputable unofficial tier of medium.

Issues (3)

E004
CRITICAL

Prompt injection detected in skill instructions.

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

W012
MEDIUM

Unverifiable external dependency detected (runtime URL that controls agent).

Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 19, 2026, 11:12 AM
Issues
3