The Agent Skills Directory

[SAFE]: The skill provides a structured markdown template and instructions for an agent to document its work for continuation by another agent. It is a legitimate utility for maintaining context across sessions.
[NO_CODE]: The skill does not contain any executable scripts, shell commands, or external dependencies. It relies entirely on natural language instructions.
[PROMPT_INJECTION]: The skill establishes a protocol for inter-agent communication, which is a potential surface for indirect prompt injection. 1. Ingestion points: The handoff document is processed by subsequent agents. 2. Boundary markers: The structure uses standard markdown headers. 3. Capability inventory: The skill itself lacks any tools for file system modification, command execution, or network access. 4. Sanitization: No input validation or output sanitization is performed, as the skill merely formats existing session data.

handoff-work