oma-dev-workflow

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

• Third-party content exposure detected (high risk: 0.80). The skill's release and CI workflows (see resources/release-coordination.md and SKILL.md) explicitly run GitHub CLI commands (e.g., bunx gh pr list, bunx gh release list / view, bunx gh run list) and instruct checking PRs, changelogs and diffs — i.e., it ingests user-generated, public GitHub content that the agent is expected to read and that can materially influence actions like merging or triggering releases.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill includes a runtime install command that runs remote shell code: "curl https://mise.run | sh" — this fetches and executes code from mise.run and is presented as the required installation step for the skill, so it directly executes remote code at runtime.