The Breadcrumb skill structures handoffs for multi-iteration agent work. It instructs an agent to read a breadcrumb directory (readme.md plus previous NNN-progress.md files), make progress on the task autonomously, and write a new numbered progress file capturing what was done, remaining work, and important context. This guarantees a durable, machine-detectable trail so subsequent agents can resume without lost context.
Use Breadcrumb when a task is too large or uncertain to finish in one run and requires iterative progress across multiple agents or sessions. Typical scenarios: long refactors, research tasks, multi-step content creation, or multi-agent orchestration where state must be preserved between runs.
Agents that can read and write repository files and commit changes (code-aware assistants, repo-integrated agents). This skill is repo-agnostic and focuses on process rather than specific tooling.
This skill has not been reviewed by our automated audit pipeline yet.
