v4.2.0

[4.2.0] - 2026-06-10 - "Mission Control Handoffs"

Added

• Added missionControl.executionPlan to projscan start / MCP projscan_start: a phased, machine-readable plan with ready commands, blocked inputs, follow-up steps, proof commands, and done criteria derived from Mission Control.
• Added missionControl.runbook, a compact Markdown handoff with ready commands, blocked inputs, proof commands, and done criteria for agents or humans continuing the task.
• Added step dependency metadata (dependsOn, blockedBy, unlocks) to Mission Control execution-plan steps so agents can tell which ready command or input unlocks each blocked follow-up.
• Added missionControl.executionPlan.cursor, a direct pointer to the next ready command, blocked input, proof command, or done criterion so agents do not need to traverse the full phase graph.
• Added the execution cursor to missionControl.handoff.currentStep and missionControl.runbook.currentStep, plus a Current Cursor section in Markdown runbooks.
• Added missionControl.resume, copied into handoff and runbook payloads, with a runnable command block or instruction plus a compact resume prompt.
• Added resolved resume.unlocks / resume.blockedBy references so resumed agents can see labels and instructions for the steps unlocked or blocking the current cursor.
• Added a resume-aware missionControl.handoffPrompt so the smallest copyable handoff starts from the current cursor and includes labeled unlocks, blockers, done criteria, and ready proof.
• Added optional missionControl.resume.toolCall so MCP agents can call the current cursor step directly without reverse-engineering CLI commands.
• Added optional missionControl.resume.followUps so resumed agents can see the placeholder follow-up templates unlocked by the current step without traversing the full execution plan.
• Added optional missionControl.resume.inputBindings so resumed agents can map placeholders like <symbol-from-search> to unlocked input steps before calling follow-up templates.
• Added optional missionControl.resume.checklist, an ordered resume task card with the current command, inputs, follow-ups, remaining proof commands, and done criteria.
• Added optional missionControl.resume.remainingProofItems, a complete ordered proof queue with step ids, status, commands, and optional MCP tool calls.
• Added optional missionControl.resume.remainingProofCommands so handoff prompts and agents can run proof that remains after the current resume command without duplicating work.
• Added optional missionControl.resume.remainingProofToolCalls so MCP agents can run mapped remaining proof steps directly without reverse-engineering CLI commands.
• Added missionControl.taskCard, exposing the paste-ready Mission Task Card Markdown to JSON and MCP clients.
• Added missionControl.reviewGate, a structured stop-and-review boundary with approval checklist, evidence commands, and Markdown for JSON and MCP clients.
• Added missionControl.reviewGate.worktree, carrying current worktree availability, changed-file count, base ref, visible files, and a human summary into review-gate handoffs.
• Added missionControl.reviewGate.proof, carrying the remaining proof queue into stop-and-review handoffs.
• Added missionControl.reviewGate.doneWhen, carrying mission success criteria into stop-and-review handoffs.
• Added missionControl.reviewGate.decisions, a structured reviewer decision menu for approving another slice, requesting changes, or reviewing a version candidate without publishing.
• Added copyable reply text to missionControl.reviewGate.decisions so reviewers can approve another slice, request changes, or request a version-candidate review without inventing permission wording.
• Added reviewer replies to the default projscan start review-gate output and saved mission bundle README.md files, so first-open review surfaces show the copyable approval text.
• Added the Mission Review Gate stop condition and reviewer reply choices to missionControl.handoffPrompt, --handoff-prompt, and saved handoff-prompt.txt so copy-only handoffs keep the no-release boundary.
• Added projscan start --review-replies and saved review-replies.txt mission bundle files so reviewers can copy approval or change-request replies without opening Markdown or JSON.
• Added projscan start --review-gate-json and saved review-gate.json mission bundle files so agents can read the full review packet without parsing full handoff JSON.
• Added missionControl.reviewGate.policy, a machine-readable approval boundary listing actions blocked until review: another slice, release, publish, deploy, push, merge, and version bump.
• Added projscan start --review-policy and saved review-policy.json mission bundle files so agents can read the review approval boundary without parsing full handoff JSON.
• Added projscan start --shortcuts-json and saved shortcuts.json mission bundle files so agents can discover shortcut commands without parsing console text.
• Added projscan start --shortcuts, a console index of copyable Mission Control shortcut commands for the current intent.
• Added projscan start --mission-script and saved mission.sh mission bundle files for running the current cursor command and remaining proof queue from one shell script.
• Added saved mission bundle proof-logs/status.jsonl; saved mission.sh writes current-command and proof-command logs plus exit-code rows under proof-logs/.
• Added saved mission bundle proof-logs/run-report.md; saved mission.sh now refreshes a Markdown pass/fail proof report with log links and review gate commands.
• Added saved mission bundle proof-logs/summary.json; saved mission.sh now records not_run, running, passed, or failed mission state for wrappers and agents.
• Added nextAction to saved mission bundle proof-logs/summary.json so agents can read the next command without parsing status.sh output.
• Added saved mission bundle status.sh, an executable status gate for proof-logs/summary.json with CI-friendly exit codes.
• Added saved mission bundle review.sh, an executable review surface that prints mission status, review gate text, run report evidence, review commands, and reviewer reply choices.
• Added quick commands to saved mission bundle README files so developers can run ./mission.sh, ./status.sh, and ./review.sh without scanning the manifest.
• Added next-action guidance to saved mission bundle status.sh output so not-run, running, failed, and passed states point to the next command.
• Added quickCommands to saved mission bundle manifests so agents can read the ./mission.sh, ./status.sh, and ./review.sh workflow without parsing README Markdown.
• Added projscan start --handoff-prompt, a console shortcut that prints only the concise Mission Control handoff prompt for piping or copy/paste.
• Added projscan start --next-command, a console shortcut that prints only the current runnable Mission Control cursor command.
• Added projscan start --next-tool-call, a console shortcut that prints only the current Mission Control cursor MCP tool call as compact JSON.
• Added projscan start --ready-tool-calls, a console shortcut that prints the current cursor and remaining MCP-callable proof queue as compact JSON.
• Added projscan start --proof-commands, a console shortcut that prints only ready Mission Control proof commands.
• Added projscan start --checklist, a console shortcut that prints only the Mission Control resume checklist.
• Added projscan start --resume-json, a console shortcut that prints only the structured Mission Control resume object.
• Added projscan start --handoff-json, a console shortcut that prints only the structured Mission Control handoff object.
• Added projscan start --save-mission <dir>, which writes a Mission Control handoff bundle with a quickstart README, next-step command and MCP-call files, copyable handoff/resume prompt files, Markdown, JSON, proof commands, and a manifest.
• Added projscan start --runbook, a console shortcut that prints only the Mission Control Markdown runbook.
• Added projscan start --task-card and task-card.md in saved mission bundles for paste-ready PR, issue, and handoff checklists.
• Added projscan start --review-gate and review-gate.md in saved mission bundles so agents can finish the current mission, report proof, and wait for approval before another slice, release, publish, or deploy.
• Added the same review gate to missionControl.handoff.reviewGate, --handoff-json, saved handoff.json, and default console output.

Changed

• projscan start --intent "<goal>" console output now includes an Execution Plan section so humans can scan the same phase model agents receive over MCP.
• projscan start --intent "<goal>" console output now includes a Run Cursor section with the single current step, its command or input instruction, and any blockers or unlocks.
• missionControl.executionPlan.currentPhase now matches the execution cursor phase so agents do not see next_action while the runnable cursor is in ready_now.
• missionControl.executionPlan.cursor now includes MCP tool / args metadata when the current cursor step is callable, and console/runbook cursor sections render that MCP call.
• missionControl.executionPlan proof steps and missionControl.resume.checklist run_proof rows now include MCP tool / args metadata when a proof command is MCP-callable.
• missionControl.handoff.readyProof.commands now uses the resume remaining-proof queue when available, avoiding duplicate proof of the current cursor command.
• missionControl.handoff.readyProof.toolCalls now carries MCP-native remaining proof calls when available.
• missionControl.handoff.readyProof.items now mirrors the complete remaining-proof item queue, including CLI-only proof steps.
• Default console Ready Proof now renders the resume-aware remaining proof commands so it does not repeat the current cursor command.
• projscan start --intent "<goal>" console output now renders Proof Queue under Ready Proof, showing MCP-callable proof steps and CLI-only proof steps without requiring --include-handoff.
• projscan start --intent "<goal>" console output now renders a Resume Checklist task card with MCP-call annotations and CLI-only proof markers without requiring --include-handoff.
• projscan start --intent "<goal>" console output now renders the concise Handoff Prompt without requiring JSON or --include-handoff.
• Mission runbooks now render the concise handoff prompt as a dedicated ## Handoff Prompt section.
• Mission task cards and runbooks now render the review-gate decision menu, so paste-ready handoffs show how a reviewer can approve another slice, request changes, or review a version candidate without publishing.
• Mission runbooks now render a Proof queue from remaining proof items, marking each step as MCP-callable or CLI only.
• Mission runbook Resume checklist rows now render inline MCP call annotations for callable checklist items and CLI only for unmapped proof commands.
• Mission task cards and runbooks now render a Review Gate section with approval and working-tree evidence steps.
• projscan start --include-handoff now prints an Agent Runbook section for the full Markdown handoff.

Fixed

• Escaped $ and backtick characters in generated projscan start shell commands so copied commands and saved mission.sh bundles treat freeform intent text as literal arguments.