Relay

Drive Relay items end-to-end through the full code pipeline (analyze → plan/superplan → review → implement → verify → resolve) without per-skill prompting. Spawns one isolated agent per item that runs the entire pipeline and returns a structured summary; the main session never absorbs analyze landscape scans, plan code blocks, or verify diffs. Default mode picks the next item per priority rules and pauses after each; sweep mode walks N items or the whole backlog. Resumable across context compaction via on-disk session state. Use when you want to automate the code pipeline and just let Relay run.

Navigate the Relay workflow. Analyzes current project state, helps you decide what to do, and routes you to the right skill. Use when you need guidance, want to file something but are not sure if it is an issue or feature, or say what should I do next.

Set up the Relay workflow system for a new project. Creates directory structure, status files, and runs project-specific configuration. Use once per project.

Analyze and validate an item before implementation. Checks if the problem still exists, identifies root cause, and maps blast radius. Use as the first step when starting work on an item from relay-ordering.md.

Generate prioritized work ordering in relay-ordering.md. Analyzes dependencies, severity, and complexity to produce phased implementation plan. Use after relay-scan updates status.

Create a detailed, step-by-step implementation plan with risk mitigation. Use after relay-analyze has validated the item and mapped the blast radius.

Close out and archive completed work. Creates implementation docs, archives resolved items, and updates ordering. Use after relay-notebook or relay-verify.

Scan the project documentation and codebase to generate relay-status.md with current state of all tracked items. Use periodically or after creating/resolving issues and features.

Creates a detailed, step-by-step implementation plan with risk mitigation. Use after relay-analyze has validated the item and mapped the blast radius. Spawns 5 competing agents with distinct strategies, once complete, synthesizes the best approach for the agents outputs. Use instead of relay-plan when you want broader exploration of implementation options.

Verify implementation against the finalized plan. Checks diffs, completeness, correctness, and runs regression tests. Use after code changes are implemented.

Adversarial review of the implementation plan. Tries to break the plan by finding holes, edge cases, and regressions. Final gate before writing code. Use after relay-plan or relay-superplan.

Walk an entire active exercise session end-to-end without per-item prompting. Spawns one isolated agent per work item that runs /relay-exercise-run then /relay-exercise-file with auto-decisions, and reports a summary back. Default mode walks all `mapped` capabilities in `_control.md` order; goal mode walks all non-terminal Journey steps in step order. Use when you want to sweep a session end-to-end and protect the main session's context window.

Map project capabilities for stress-testing. Default mode (no argument) scans bottom-up; goal mode (/relay-exercise "<goal>") builds a top-down journey with gap discovery. Creates a session subfolder under .relay/exercise/<session>/ with _control.md (Journey for goal mode, Context Chains for default) and updates the master registry at .relay/relay-exercise.md. Use at project onboarding, when the project has evolved, or to stress-test whether a specific goal is achievable.

Archive abandoned brainstorm files in .relay/features/. Use periodically to clean up brainstorms that were started but never completed through relay-design.

Walk exercise findings within the active session with the user, file as issues or brainstorm seeds with session-aware Source headers, and update both the session _control.md and the master hub. Use after /relay-exercise-run produces findings.

Execute realistic scenarios against a capability (or group) within the active session. In goal-mode sessions, walks the journey end-to-end with adaptive gap handling (alternative / file / skip). Writes to .relay/exercise/<session>/<capability>.md (default) or step-<N>-<capability>.md (goal). Use after /relay-exercise creates a session.

Interactive brainstorming session for a new feature idea. Explores approaches, trade-offs, and produces a brainstorm file. Use when you have a feature idea to explore.

Scan the codebase to discover bugs, gaps, inconsistencies, and technical debt. Creates issue files in .relay/issues/. Use periodically or after significant code changes.

Design features from a brainstorm file into detailed specs. Creates individual feature files with architecture, interfaces, and data flow. Use after relay-brainstorm produces a READY FOR DESIGN brainstorm.

File a specific bug or gap as an issue. Redirects features to relay-brainstorm. Use when you have identified a specific problem to document.

Create and run a Jupyter verification notebook that exercises the real project API end-to-end. Use after relay-verify confirms implementation is complete.