marshal

Multi-step code investigation and fixes are automated by orchestrating skills and contextual project data into a single-session plan. This removes manual context gathering and reporting coordination to execute multi-step developer workflows quickly.

Orchestrating cross-file analysis and audits in one session requires chaining existing skills and tools to gather repository context. This approach reads project conventions, identifies scope, and produces structured findings with file:line citations.

Automated code review and audits require access to repository files and the ability to load sub-skills. This access allows the orchestrator to gather project context, enforce agent timeouts, and apply fixes within the session.

Multi-step agent orchestration handles execution timeouts by enforcing agent limits with automatic retries. This robust execution model ensures workflows complete successfully and produce structured reports with a HANDOFF summary.

You can use an orchestrator to assess an authentication module, identify security or UX gaps, and apply fixes if feasible. It then produces a concise Marshal Report with next steps and structured findings.

Single-session multi-step code fixes are limited to investigations and audits that can be bounded within one execution. Workflows requiring cross-session continuity must rely on the HANDOFF summary to resume work.