The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow where agents contribute to shared decision logs within the .squad/ directory. Because agents are instructed to read these logs to gain context, this creates a potential surface for indirect prompt injection where instructions authored by one agent could influence the behavior of others.
Ingestion points: Agents are instructed to read decisions.md and their history.md (Context section).
Boundary markers: No specific delimiters or safety warnings are provided for the content of decision logs to prevent agents from obeying embedded instructions.
Capability inventory: The skill explicitly grants file-write access to the .squad/decisions/inbox/ directory for all agents.
Sanitization: The skill does not provide instructions for sanitizing or escaping content written to shared logs.
[COMMAND_EXECUTION]: The instructions include a fallback command git rev-parse --show-toplevel to identify the project root. This is a standard utility command used in development environments.

agent-collaboration