Planning & Specification Workflows

Executes an approved agentmaster-plan plan and chains straight into the full code review. Use when the user has accepted a plan and wants it built — "execute the plan", "run the plan", "build it", or /agentmaster-execute. Reads the plan, resolves blocking open questions with the user, dispatches one implementer subagent per conflict-free parallel group, gates on per-task verification, then invokes agentmaster-review, which performs the frontier code review and owns the fix loop.

Cost-tiered planning pipeline. Use when the user asks to plan a feature, refactor, migration, or fix with agentmaster, or invokes /agentmaster-plan. A frontier orchestrator dispatches cheap scout/code-analyst workers for evidence, detects the project toolchain, drafts a parallel-safe plan with an evidence ledger, and survives adversarial critique. Supports --lite and --headless.

Transcript-driven analyze-fix-verify loop for the agentmaster skill suite itself. Use after a pipeline run leaves artifacts behind — .transcripts/ snapshots, root-level run transcripts (OUTPUT.md-style), generated docs, or accumulated telemetry — when the user wants the suite's own skills to keep improving from real usage, or invokes /agentmaster-retro. Grades corpus artifacts against a rubric with code-analyst (haiku scout inventories, sonnet code-analyst grades ALREADY-FIXED/PARTIALLY-FIXED/UNFIXED against current skill text), ranks weaknesses, dispatches implementer fixes, verifies with the plan-structure lint and full gate, and writes a dated retro to .agentmaster/retro/ so the next run closes the loop.

Cost-tiered adversarial code review. Use after implementer subagents complete a agentmaster-plan plan (the plan's review gate invokes this), or whenever the user wants a rigorous review of a diff, branch, PR, or recent changes. Works from the assumption that the code is bad — poorly implemented, doesn't scale, violates YAGNI, SOLID, or DRY, is hard to test, badly structured, or insecure — and makes it prove otherwise with evidence. Keeps frontier reasoning for adjudication only: dispatches scout (haiku) and code-analyst (sonnet) for all evidence, then approves or emits fix tasks and dispatches implementers. Trigger on "review the changes", "review this diff", "review this branch", or as the final task of an executed plan.