Autonomous feature development — current checkout

Implement this feature idea directly in the user's current Git checkout:

$ARGUMENTS

This skill is the current-checkout counterpart to autonomous-feature. The user has already created a clean non-main/non-master feature branch and wants the agent's edits to land directly in that branch — no disposable worktree, no .claude/worktrees/* clone, no extra worktree-* branch. Use ultrathink for architecture, compatibility, and review triage.

Non-negotiable boundaries

• This skill must NOT call EnterWorktree / ExitWorktree. All edits land in the current checkout. Do not create or enter .claude/worktrees/*.
• Refuse to run on main or master. If the current branch is one of those, stop and ask the user to create a feature branch first (or use /autonomous-development:autonomous-main when they have explicitly authorized direct edits on main).
• Refuse to run on a dirty working tree. git status --porcelain must be empty before initializing. If it is not, stop and report which entries are dirty.
• Do not create commits. The user will review and commit with their normal git diff/commit flow.
• Preserve unrelated user changes.
• Never push, merge, publish, deploy, rotate credentials, or modify remote infrastructure.
• Never use danger-full-access, --yolo, bypassPermissions, or equivalent unrestricted modes.
• Never apply an irreversible database migration or delete user data.
• Do not weaken authorization, validation, tests, or static checks to make the workflow pass.
• Codex planning and review executions must remain read-only.
• Use no more than the configured review-round budget.
• Treat every Codex finding as a proposal requiring evidence-based triage.

Driver loop

1. Confirm this is a Git repository. Inspect CLAUDE.md, repository instructions, architecture, status, and tests. Verify the current branch is NOT main/master and that the working tree is clean before initializing — the controller will fail closed on both, but checking up front produces a clearer message for the user.

2. Initialize in current-checkout mode (no worktree):

   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/controller.py" doctor
   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/controller.py" init \
     --feature "$ARGUMENTS" \
     --mode standard \
     --worktree-mode current

   init prints the run-state.json path and run ID. With multiple concurrent runs, pass --run-id <run-id> to all subsequent commands. If doctor reports a missing prerequisite, mark the run blocked and report it rather than bypassing it.

3. Repeatedly ask the controller for the next phase, then execute it:

   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/controller.py" next-action --json

   The response gives phase, required_action, completion_condition, and references. Read the referenced file under ${CLAUDE_PLUGIN_ROOT}/skills/autonomous-feature/references/ for that phase and follow it until the completion condition holds.

4. Do not declare success until evaluate succeeds:

   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/controller.py" evaluate
   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/controller.py" status
   python3 "${CLAUDE_PLUGIN_ROOT}/scripts/controller.py" usage-report

Final report

• the implemented behavior;
• principal files changed (visible to the user via plain git diff);
• verification commands and results;
• Codex review rounds and disposition of findings;
• adversarial review result when one was required;
• per-phase usage table from usage-report;
• remaining risks or explicit blocked reason;
• a suggested conventional commit message (the user commits manually — this skill must not commit).