create-handoff

Pass

Audited by Gen Agent Trust Hub on Sep 23, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill establishes a mechanism for agents to pass executable context to future sessions. These handoff documents include 'Resume contracts' with commands such as 'Next step' and 'Re-arm', which a resuming agent is instructed to follow. This introduces a surface where instructions from one session can influence the actions of a subsequent session.
  • Ingestion points: Handoff documents are written to the thoughts/shared/handoffs/ directory and are consumed by the resume-handoff skill.
  • Boundary markers: The skill uses a defined YAML frontmatter and Markdown structure but lacks explicit delimiters or instructions to prevent the processing of malicious content within the fields.
  • Capability inventory: The skill utilizes Bash, Write, and Read. The resuming agent is encouraged to execute shell commands provided in the handoff.
  • Sanitization: While file paths are sanitized, the natural language content and commands generated from the session context are not validated for safety.
  • [COMMAND_EXECUTION]: The skill uses shell commands to manage files, resolve paths, and interact with version control systems.
  • Evidence: The skill runs git to collect metadata and uses a local helper script handoff-durability.sh for atomic file operations. It also invokes the humanlayer CLI to perform synchronization operations, which is consistent with the vendor's provided infrastructure.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 23, 2026, 08:37 PM
Security Audit — agent-trust-hub — create-handoff