run-deep-swe

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). These are GitHub repositories (git clone / git+https install sources) from an organization that is not a well-known vendor in the doc and instructs running code/containers directly, which makes them potentially risky as unverified sources for executable code.

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

• Third-party content exposure detected (high risk: 0.85). The required workflow runs pier run ... --agent mini-swe-agent --model ... --model-class openrouter, which causes the agent to call OpenRouter at runtime and ingest the model’s responses (outsider-authored free text) into the LLM context for subsequent reasoning.

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

• Potentially malicious external URL detected (high risk: 0.90). The instructions tell the user to git clone https://github.com/datacurve-ai/deep-swe (and optionally install git+https://github.com/datacurve-ai/pier), and the cloned repository provides task definitions and code that pier executes to drive the agent's prompts/instructions at runtime, so these remote URLs directly supply required executable content and prompts.