handoff
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from local project files, specifically
CLAUDE.md,AGENTS.md, and previous handoff files. This creates a surface for indirect prompt injection if those files contain malicious instructions meant to influence the handoff summary. - Ingestion points: The skill reads
CLAUDE.md,AGENTS.md, and existing handoff files during the 'Procedure' phase. - Boundary markers: The skill does not use explicit boundary markers for ingested data but includes a 'Core Principle' to "Trust nothing blindly" and generates a warning prompt for the next agent to verify all claims against the actual source code.
- Capability inventory: The skill uses file-writing capabilities to save the handoff to the temporary directory or project root.
- Sanitization: The skill explicitly instructs the agent to redact secrets, API keys, and PII from the output.
- [COMMAND_EXECUTION]: The skill instructs the agent to perform file system write operations to save the handoff content to the OS temporary directory (e.g.,
$TMPDIR) or the project root (HANDOFF.md). This is the intended primary function of the skill to persist session state between different agent contexts.
Audit Metadata