gitcode-issue-reply

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text from the GitCode issue/comments authored by other users is fetched via GitCode API in scripts/prepare_issue_reply.py (fetch_issue/fetch_comments using api_get), then embedded into the LLM prompt as issue_content_plain/prompt_draft_partial (and image URLs are also derived from that text).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill makes runtime requests to DeepWiki's MCP endpoint (https://mcp.deepwiki.com/mcp) to fetch knowledge text that is inserted into the prompt template and it also downloads arbitrary image URLs found in issue bodies (e.g. https://.../image.png) during runtime and incorporates their analysis into the generated reply, so external content can directly influence the agent's prompts.

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

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