codealive-context-engine

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt includes an explicit "not recommended" option that tells the agent to accept a pasted API key in chat and save it via a command like python setup.py --key THE_KEY, which requires the LLM to handle and emit the secret verbatim (exposing it in chat/commands).

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and ingests code from external/untrusted repositories via the CodeAlive API (see search.py's --include-content option and explore.py which calls client.search(include_content=True) and then client.chat with data_sources), and those third‑party code contents are synthesized by chat.py/explore.py into actionable recommendations—so untrusted repo/web content can materially influence agent behavior.