handoff

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). This skill explicitly requires reproducing verbatim user messages, file contents, and command outputs (Blocks 3, 4, and 9), so any API keys, tokens, or passwords present in those sources would be emitted by the LLM, creating a high exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.85). In both prepare-mode and resume-mode, the skill reads/writes the HANDOFF Markdown file under .claude/handoffs/ and includes Block 4 “All user messages” verbatim; those user messages are outsider-authored free text (not chosen by the operating user to be introduced as trusted context), which becomes LLM-readable prose in the agent’s context when the file is read in resume mode.

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

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