semble

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

• Third-party content exposure detected (medium risk: 0.65). Outsider-authored free text can enter the agent’s LLM context when the skill uses semble search/MCP to index and return code chunks from a user-supplied repo URL or other non-user-controlled source (e.g., https://github.com/...), where the returned snippet content is read by the agent as prose.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly supports indexing/searching remote git repositories at runtime (e.g., "semble search ... https://github.com/MinishLab/model2vec" and SembleIndex.from_git("https://github.com/MinishLab/model2vec")), and the fetched code chunks are injected into agent prompts, so external repo URLs can directly control the agent's prompt/context.

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

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