controller-gated-mpd
Use when multiple branches, worktrees, or pull requests must converge through a controlled integration wave with frozen-base, exact-head evidence, READY…
Skill by eyildirim82
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/eyildirim82-controller-gated-mpd
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Python package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/controller-gated-mpd
What's inside
- Skill name
controller-gated-mpd- Description
- Use when multiple branches, worktrees, or pull requests must converge through a controlled integration wave with frozen-base, exact-head evidence, READY binding, or serialized merge-train requirements.
- License
- MIT
- Instructions
- 68 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (23)
- README.md
- SKILL.md
- openai.yaml · agents/
- evals.json · evals/
- controller-review-template.md · references/
- coordinator-prompt-template.md · references/
- dispatch-modes.md · references/
- project-profile-template.md · references/
- wave-profile-template.json · references/
- worker-prompt-template.md · references/
- worker-result-contract.md · references/
- actions_metadata.py · scripts/
- approve_ready.py · scripts/
- assert_frozen_base.sh · scripts/
- check_deps.sh · scripts/
- check_exact_head.py · scripts/
- check_trunk_evidence.py · scripts/
- merge_train.sh · scripts/
- policy.py · scripts/
- render_handoff_prompts.py · scripts/
- validate_wave_profile.py · scripts/
- verify_ready.py · scripts/
- wave_state.py · scripts/
SKILL.md (first lines)
# Controller-Gated MPD Controller-Gated MPD is the Adaptive Superpowers execution engine for distributed multi-lane integration. It is **not** a second orchestrator. ## Authority split Adaptive Superpowers owns: - FAST / STANDARD / CRITICAL classification; - inherited risk floors and escalation; - effect authorization; - TDD/review/verification strength; - whether parallelism is worthwhile. Controller-Gated MPD owns: - frozen-base and lane ownership contracts; - wave/dependency state; - prompt/native transport for worker execution; - exact-head GitHub Actions metadata evidence; - READY@SHA materialization and invalidation; - serialized merge-train mechanics; - post-merge trunk proof. ## Activation Use MPD only when work is a real distributed integration problem: multiple isolated branches/worktrees/PRs must converge to one trunk, especially when frozen-base, ownership, dependency DAGs, reserved migrations/resources, exact-head CI, or controlled merge order matter. Do not activate MPD merely because two subtasks exist. Ordinary parallel work stays with `dispatching-parallel-agents` / `subagent-driven-development`. An MPD wave has a minimum **STANDARD** coordination floor. Child lanes inherit the wave floor and may escalate to CRITICAL, never downgrade below the floor. ## Transport - `native-dispatch`: use when the runtime can actually start isolated workers. - `prompt-handoff`: use in a normal ChatGPT conversation without native worker dispatch. For **Prompt-Handoff Parallelism**, generate complete zero-context prompts for the coordinator and currently dispatchable lanes. Tell the user to open worker prompts in separate ChatGPT windows. Never claim workers were launched when only prompts were generated. Blocked downstream lane prompts are withheld by default. A preview may be rendered only when explicitly marked `DO NOT START` with unmet prerequisites. ## Evidence and READY Worker prose, pasted logs, screenshots, and worker-reported SHAs are handoff hints only. They are **not evidence** and must not create READY. The coordinator independently re-reads Git/PR/Actions state. READY is a machine integration capability bound to the exact repo, PR, head, base, profile digest, effective Adaptive risk, run ID/attempt, and evidence digest. READY is not user approval. V5 schema/protocol authority does not upgrade into V6 authority. V6 fails closed on legacy READY/evidence. ## Merge train Plan mode is non-mutating. A mutating merge train requires Adaptive authorization `merge == granted`. `unknown` or `denied` may preview the train and reach READY, but must stop before merge mutation. Already-granted merge authority is reused without asking again. ## Ledger While MPD is active, `.superpowers/mpd/<wave>/` is the canonical recovery ledger. Do not create a competing generic SDD ledger for the same wave. ## References - `references/dispatch-modes.md` - `references/project-profile-template.md` - `references/wave-profile-template.json` - `references/worker-prompt-template.md` - `references/coordinator-prompt-template.md` - `references/worker-result-contract.md` - `references/controller-review-template.md`
From eyildirim82/adaptive-superpowers, MIT.
More from eyildirim82/adaptive-superpowers
- 0.529score
subagent-driven-development
eyildirim82
Use when an implementation plan contains independent coherent tasks or lanes that benefit from isolated subagent contexts.
SkillBFreshness: BClaude Code +9 - 0.529score
brainstorming
eyildirim82
Use when a task needs product/design exploration before implementation, especially new behavior, new subsystems, or unresolved choices.
SkillBFreshness: BClaude Code +9 - 0.528score
using-superpowers
eyildirim82
Use when starting a coding or software-development task that may need Superpowers process or domain skills.
SkillBFreshness: BClaude Code +9 - 0.523score
using-git-worktrees
eyildirim82
Use when the active execution profile requires or benefits from an isolated Git workspace.
SkillBFreshness: BClaude Code +9 - 0.523score
systematic-debugging
eyildirim82
Use when a bug, failing test, build error, integration issue, or unexpected behavior needs diagnosis before a fix.
SkillBFreshness: BClaude Code +9 - 0.523score
finishing-a-development-branch
eyildirim82
Use when implementation is ready for final verification and any authorized Git integration or cleanup effect.
SkillBFreshness: BClaude Code +9 - 0.523score
requesting-code-review
eyildirim82
Use when a change needs an independent engineering review based on diff consequence, risk, or integration boundary.
SkillBFreshness: BClaude Code +9 - 0.523score
writing-plans
eyildirim82
Use when implementation needs an explicit multi-step plan or worker contracts before code changes.
SkillBFreshness: BClaude Code +9 - 0.523score
writing-skills
eyildirim82
Use when creating, editing, or verifying reusable agent skills and process documentation before distribution.
SkillBFreshness: BClaude Code +9 - 0.523score
executing-plans
eyildirim82
Use when a written implementation plan should be carried through in the current or separate execution session.
SkillBFreshness: BClaude Code +9 - 0.523score
dispatching-parallel-agents
eyildirim82
Use when multiple independent work domains could be executed concurrently without shared-state conflicts.
SkillBFreshness: BClaude Code +9 - 0.519score
receiving-code-review
eyildirim82
Use when code review feedback must be evaluated, especially when findings are ambiguous, risky, or technically debatable.
SkillBFreshness: BClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 21
- Last commit (this folder)
- Sep 13, 2026 (1 month ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from eyildirim82
- 0.529score
brainstorming
eyildirim82
Use when a task needs product/design exploration before implementation, especially new behavior, new subsystems, or unresolved choices.
SkillBFreshness: BClaude Code +9 - 0.529score
subagent-driven-development
eyildirim82
Use when an implementation plan contains independent coherent tasks or lanes that benefit from isolated subagent contexts.
SkillBFreshness: BClaude Code +9 - 0.528score
using-superpowers
eyildirim82
Use when starting a coding or software-development task that may need Superpowers process or domain skills.
SkillBFreshness: BClaude Code +9 - 0.523score
finishing-a-development-branch
eyildirim82
Use when implementation is ready for final verification and any authorized Git integration or cleanup effect.
SkillBFreshness: BClaude Code +9 - 0.523score
using-git-worktrees
eyildirim82
Use when the active execution profile requires or benefits from an isolated Git workspace.
SkillBFreshness: BClaude Code +9 - 0.523score
requesting-code-review
eyildirim82
Use when a change needs an independent engineering review based on diff consequence, risk, or integration boundary.
SkillBFreshness: BClaude Code +9
Cite this page
controller-gated-mpd skill. Agent Mart. https://agentmart.dev/skills/eyildirim82-controller-gated-mpd (last updated 2026-10-24)