throughline-consolidate

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill instructs the agent to read session logs and include "a quote or close paraphrase" from each log as evidence, which would force the LLM to reproduce verbatim content it reads — so if any logs contain API keys, tokens, passwords, or other secrets, the agent would be required to output them directly, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.75). This skill reads and mines outsider-authored session log markdown files from DATA/logs/handoff-*.md (and potentially .agent/handoff/logs/*.md / .agent/handoffs/*.md), extracting quoted/paraphrased text into the agent’s LLM context during Phase 2.