orca-onboard

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The skill aggressively sends project data to external LLM services (opencode/DeepSeek, Claude), clones/copies a remote "setup" repo and can install a global "harness" (hooks) and instructs pushing local SKILL.md upstream — behaviors that enable data exfiltration and potential supply‑chain/backdoor persistence.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill runs git clone https://github.com/rheinmir/setup.git (and later references git@github.com:rheinmir/setup.git) at runtime and then executes scripts from /tmp/orca-llmwiki-bootstrap (e.g., bash /tmp/orca-llmwiki-bootstrap/harness/scripts/install-harness.sh), so it fetches and executes remote code which the onboarding relies on to bootstrap llmwiki/harness.

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)