handoff

Pass

Audited by Gen Agent Trust Hub on Mar 2, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill performs local filesystem operations to save handoff documents. Specifically, it writes markdown files to the .claude/handoffs/ directory.
  • [DATA_EXFILTRATION]: The skill aggregates session data, including user context and code changes, for the purpose of local persistence. No network activity or external data transmission was detected.
  • [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because it stores session content for later consumption by an agent.
  • Ingestion points: The skill reads session history, user inputs, and modified source code files (SKILL.md).
  • Boundary markers: The output uses standard Markdown headers (e.g., ## Decisions Made) but does not include explicit delimiters or instructions telling the future agent to ignore instructions embedded within the handoff data.
  • Capability inventory: The skill possesses file-writing capabilities to the local .claude directory.
  • Sanitization: There is no evidence of content sanitization or filtering to remove potential instructions from the captured session context.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 2, 2026, 04:59 PM