erk-planning

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly instructs fetching and parsing GitHub PR comments (e.g., gh pr view 123 --comments --json comments in SKILL.md and references/workflow.md) to extract the plan-body from the first comment, which is untrusted user-generated content that the agent reads and acts on to update/implement plans.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly fetches PR content at runtime (e.g., via "gh pr view 123 --comments" and by using the PR URL "https://github.com/owner/repo/pull/123"), and that fetched comment's plan-body is parsed and used as the plan content that drives agent behavior—so external GitHub content can directly control the agent's prompts/instructions.