release-notes

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

• Third-party content exposure detected (high risk: 0.90). The skill instructs the agent to fetch and parse public GitHub content (e.g., curl raw.githubusercontent.com to read install.sh for LATEST_STABLE_AGENTSTACK_VERSION and to run gh pr view / gh pr list to read PRs and PR comments), which are untrusted, user-generated third‑party sources that the agent is expected to read and use to guide decisions about which changes are high-impact.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill instructs runtime fetching of https://raw.githubusercontent.com/i-am-bee/agentstack/install/install.sh via curl to extract LATEST_STABLE_AGENTSTACK_VERSION, which the skill uses to determine the release scope (i.e., it is fetched at runtime, its contents directly influence the agent's behavior/decisions, and the workflow depends on it), so it meets the criteria for a risky external dependency.