The Agent Skills Directory

[PROMPT_INJECTION]: The skill's primary purpose is to generate a 'Start Prompt' for next-thread agents. This creates a surface for indirect prompt injection where malicious instructions embedded in the goal, decisions, or risk context are persisted into a file and subsequently processed as instructions by another agent.
Ingestion points: The workflow defined in SKILL.md and the arguments passed to scripts/create_handoff.py (goal, title, context).
Boundary markers: The generated markdown uses standard headers but lacks explicit delimiters or instructions to ignore embedded commands within user-provided fields.
Capability inventory: The skill utilizes scripts/create_handoff.py to write files to the local filesystem.
Sanitization: There is no sanitization or escaping of input data before it is written to the markdown handoff packet.
[COMMAND_EXECUTION]: The script scripts/create_handoff.py uses the pathlib and argparse libraries to create directories and write files based on user-supplied arguments. While this is the intended functionality for saving handoff skeletons, the lack of path validation on the --output argument could allow for writing files to unexpected locations within the agent's filesystem permissions.

agent-handoff