The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill establishes an orchestrator-worker pattern where data from untrusted or external sources is integrated into the agent's context.
Ingestion points: The agent is instructed to read durable records from GitHub Issues or ISSUES.md, and synthesize results from worker-generated files located in the scratchpad/ directory.
Boundary markers: There are no instructions or delimiters provided to help the agent distinguish between its own operational instructions and potentially malicious instructions embedded in the task descriptions or worker outputs.
Capability inventory: The skill relies on shell command execution (e.g., cat, ls, find, echo), file system read/write access to project directories, and Git operations (SKILL.md).
Sanitization: The instructions lack any requirement for sanitizing, validating, or escaping content retrieved from external trackers or distributed worker agents before it is processed by the orchestrator.

swarm-coordination