audiocontrol

Run upfront discovery for a system-wide feature. Fans out discovery agents in parallel, synthesizes a strawman scope-manifest.yaml, writes the evidence trail to the feature docs directory.

Run targeted discovery for a mid-implementation operator complaint. Identifies all sibling surfaces consuming the same primitive/pattern; produces a widening proposal in the dispatch-wrapper-grammar format.

Use when shipping a new midi-macro-bridge version — bumping Cargo.toml, cutting a v-tag, publishing a GitHub Release, or updating the Homebrew tap.

Extend a feature's scope mid-implementation by adding new phases to the existing docs and issue plan without creating a new branch or worktree.

Explain the audiocontrol feature lifecycle and report the current state of active, in-progress, or partially defined features.

Wrap up a Codex work session by updating feature docs, recording development notes, and closing the loop on issue and hardware notes where needed.

Bootstrap a Codex session by reading the active feature workplan, latest journal notes, and relevant issue context before implementation starts.

Extend a feature's scope mid-implementation: add phases to the PRD, workplan, and create new GitHub issues. Reuses the existing worktree and branch.

Core implementation loop: select next workplan task, choose appropriate agent, delegate with full context, review output, update progress.

Mark a feature as complete: move docs to 003-COMPLETE, update ROADMAP.md, close GitHub issues. Runs on the feature branch BEFORE merge so bookkeeping is part of the PR.

Bootstrap a Claude Code session by reading the feature workplan, latest journal entry, and open issues. Reports context to the user so they can confirm the session goal.

Rehydrate context for an in-progress feature by reading the workplan, notes, and issue state, then propose the next implementation move.

Create feature infrastructure: branch, worktree, docs directory, and initial project documents derived from a feature definition.

Run the session extraction and analysis pipeline, then summarize the resulting trends and improvement signals.

Build, deploy, and verify the SCSI MIDI bridge on the target Pi, then report the resulting status.

Mark a merged feature complete by moving docs, updating roadmap state, and closing tracking issues.

Interview the user to define a feature, then write a single feature definition file that downstream setup steps can consume.

Drive the core implementation loop for a feature: pick the next workplan task, implement it, verify it, and update progress.

Create GitHub tracking issues from a completed workplan and backfill the workplan with the resulting issue links.

Review recent feature changes with a code-review mindset focused on correctness, regressions, risks, and missing tests.

Prepare a completed feature for merge by validating the workplan, running tests, and opening a pull request with a concise summary and test plan.

Remove local feature infrastructure such as worktree and branch without making assumptions about feature status.

Interview the user to define a new feature: problem, scope, approach, and tasks. Writes a feature-definition.md file that /feature-setup consumes.

Show the feature lifecycle workflow, explain each skill's role, and report current state of any active or in-progress features.

Create feature infrastructure: git branch, worktree, docs directory, and delegate documentation creation to documentation-engineer agent.

Create GitHub issues from a completed workplan: parent feature issue, implementation issues, then update workplan with links.

Prepare feature for merge: verify acceptance criteria, run tests, create PR with summary and test plan.

Bootstrap a feature-orchestrator session: read workplan, check issue status, report current state and next steps.

Delegate code review of recent changes to code-reviewer agent and report findings.

Remove local feature infrastructure: worktree and branch. Infrastructure-only, no opinion on feature status.

Full pipeline: extract session data, encrypt content, run LLM analysis, generate report.

Wrap up a Claude Code session by updating feature docs, writing a journal entry, and committing documentation changes.

Cross-compile the scsi-midi-bridge, deploy to the Pi, and verify it's running.