agent-handoff
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates by summarizing tool outputs, file changes, and agent communications, creating an ingestion surface for potentially untrusted data that could contain embedded instructions.
- Ingestion points: The skill captures "changed areas," "validation evidence," and "tool or command outcomes" as specified in the
SKILL.mdcapture instructions. - Boundary markers: The instructions mandate "Context Isolation" to prevent internal chatter from polluting contexts and enforce the use of structured JSON contracts (
a2a-task.json) when crossing role boundaries. - Capability inventory: The skill is configured with access to
write_file,delegate_task, andexecute_commandin the frontmatterallowed-tools. - Sanitization: The skill requires strict sanitization of user-facing artifacts, explicit redaction of PII, tokens, and credentials, and the removal of internal agentic metadata before output emission.
Audit Metadata