coexistence.md

Autopilot — Superpowers Coexistence

Part of Autopilot. Detail docs: Skills · Coexistence · Configuration · Installation · Architecture · Hooks

Autopilot is standalone-capable. This page covers how it coexists with the optional superpowers plugin: the layered model, the three deployment scenarios, and the migration note.

---

Coexistence Model — autopilot is standalone-capable, Superpowers is optional

Autopilot works fully without superpowers. If you also have superpowers installed, autopilot's orchestrator skills (ceo-agent, finish-flow, quality-pipeline, think-tank{,-dialectic}, dev-flow) consult .claude/dispatch-config.md to decide which methodology / reviewer / parallel dispatcher to delegate to.

autopilot:dev-flow sets session rules:
  → "When debugging, read .claude/debug-config.md for project context"
  → "Before committing, run autopilot:quality-pipeline"
  → "On session end, update project tracking"

Methodology dispatch (per .claude/dispatch-config.md):
  → Debugging:  superpowers:systematic-debugging (if installed) → autopilot:debug
  → Testing:    superpowers:test-driven-development + autopilot:test-strategy (complementary)
  → Profiling:  autopilot:profiling (no superpowers equivalent)
  → Team:       autopilot:team (allocation) + superpowers:dispatching-parallel-agents (dispatch)
  → Review:     autopilot:reviewer → superpowers:requesting-code-review (fallback)

This was historically positioned as「sets the rules; Superpowers executes」(v2.0-v2.6); v2.7.0 preserves that model when superpowers is installed while making autopilot also work as a standalone plugin. See Superpowers Coexistence for per-scenario UX.

---

Superpowers Coexistence

Autopilot supports three deployment scenarios:

A. You have superpowers installed (user-level or marketplace)

If you also have superpowers, autopilot's orchestrators delegate tactical execution to superpowers via .claude/dispatch-config.md chains.

.claude/dispatch-config.md example (paste into your project):

## Code Review
- autopilot:reviewer
- superpowers:requesting-code-review

## Parallel Dispatch
- superpowers:dispatching-parallel-agents
- native

## Methodology Preferences

### Debugging
- superpowers:systematic-debugging
- autopilot:debug

### Testing methodology
- autopilot:test-strategy
- superpowers:test-driven-development

superpowers ≥ v5.1.0 note: the standalone superpowers:code-reviewer agent was removed and folded into the requesting-code-review / receiving-code-review skills (verified against obra/superpowers v6.0.3, 2026-06). Chains above use the current name; autopilot:reviewer remains the methodology-disciplined primary, so this fallback is optional.

B. You do NOT have superpowers installed

Autopilot runs fully standalone. Orchestrators fall through to autopilot's own fallback skills (autopilot:debug, autopilot:test-strategy, autopilot:team, autopilot:profiling) and native parallel dispatch (multiple Task tool calls in one response).

One standalone gap to know: the red-green-refactor TDD coding loop has no native autopilot skill — autopilot:test-strategy is orthogonal (test pyramid / baseline / failure funnel), not a TDD substitute. For TDD specifically, install superpowers (test-driven-development) or run the red-green cycle by hand.

No .claude/dispatch-config.md needed — the defaults documented at the top of project-config-template/dispatch-config.md match this scenario.

C. You have superpowers user-level but want pure-autopilot in a specific project

Use .claude/settings.json's disabledSkills to hard-cut superpowers skills per-project:

{
  "disabledSkills": [
    "superpowers:systematic-debugging",
    "superpowers:test-driven-development",
    "superpowers:dispatching-parallel-agents",
    "superpowers:requesting-code-review"
  ]
}

This is a Claude Code native mechanism; autopilot doesn't need a config flag for it.

Migration note (v2.6.0 → v2.7.0)

If you upgrade from v2.6.0 and previously removed debug, test-strategy, team, or profiling entries from your CLAUDE.md skill routing tables (expecting these skills to remain absent), be aware they're back as fallback skills in v2.7.0 and may now trigger on the corresponding keywords. To suppress: add them to .claude/settings.json's disabledSkills.

Codex Plugin Coexistence (openai/codex-plugin-cc)

Autopilot coexists with the official OpenAI codex plugin on Claude Code hosts:

Surface	Autopilot	Codex plugin	Rule
Peer consult	dispatch-explore.sh / dispatch-author.sh (any host)	codex:codex-rescue / companion task (~9s, thread-resumable)	Plugin preferred on CC when installed; rails are the portable fallback
Code review	qc panels + dispatch-review.sh (calibrated: known-bad 10/10 bar)	/codex:review (Spark-locked, uncalibrated)	Autopilot rails stay authoritative; plugin review channel gated on calibration (BACKLOG)
Stop-time gate	.githooks/pre-push qc-gate (QC-Verdict trailer)	optional stop-review-gate hook	Keep the plugin's gate DISABLED — one authoritative gate, not two
Write/labor	dispatch-hetero.sh (worktree + cgroup + artifact verify)	rescue --write (sandbox posture unverified)	Labor stays on autopilot rails until the plugin's write path is spiked
Repo hygiene	—	plugin install writes .claude/settings.json in cwd	gitignored since v2.31.19+ (.claude/settings.json)

CodeForge status-line feed (optional, v2.36.1)

If codeforge is your Claude Code statusLine, autopilot's context gates read the real model window and per-subagent usage from the tmpfs live files it writes (hooks/README.md § Live context feed). Enable the subagent half with codeforge install --subagent-statusline. Nothing else in autopilot depends on codeforge: without it every hook keeps its v2.36.0 behaviour.

Agent Call Coexistence

Agent Call is an optional transport for already-running persistent local sessions. Autopilot keeps lifecycle and ownership policy; it does not absorb Agent Call's harness adapters.

• Claude↔Claude uses native ListAgents / SendMessage when the exact target is available.
• Other persistent local peers use autopilot:agent-call, which calls the separately installed agent-call CLI.
• A failed persistent-peer send is terminal for that route. It never falls through to Task, a foreman, dispatch-hetero.sh, or another worker-spawning rail.
• New temporary implementers/reviewers continue to use Autopilot orchestration; Agent Call does not schedule them.
• Project skills retain ownership, conflict, merge, review, and evidence rules, but should not copy tmux registration or injection recipes.