inno-code-survey

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly requires extracting and emitting complete, verbatim code snippets and a full "model_survey" report from local/cloned repositories, which would force the LLM to reproduce any hardcoded API keys, tokens, or passwords present in those files.

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

• Third-party content exposure detected (high risk: 0.95). Phase A/B both cause the agent to ingest outsider-authored free text from cloned public GitHub repositories (e.g., README.md and source files read via cat/tree/find/grep into the LLM context during repo verification and code surveying).

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

• Potentially malicious external URL detected (high risk: 1.00). The included helper script and agent workflow perform runtime calls to the GitHub Search API (https://api.github.com/search/repositories) and git-clone URLs (e.g., https://github.com//.git), fetching external repository code that is then read and injected into the Code Survey Agent's prompts/context (so external content directly controls agent inputs and is required for the skill).