CCC works with your skills.

Superpowers, gstack browse, Total Recall, Watchtower. Good skill packs already do the work inside one session. CCC does not replace them. It gives them a fleet: a persistent, visible, multi-session home. This page shows what works today, separated plainly from the roadmap.

What a skill pack owns

The work inside a session.

Superpowers plans, runs TDD, fans out subagents, reviews a diff. gstack browse drives a browser. Total Recall remembers. Each is excellent at its job, and blind to every other session on your machine.

What CCC owns

The fleet around them.

Persistent sibling sessions you can spawn, inject into, and ask, rendered on one board. CCC reads each engine's on-disk state, so a pack's subagents and a plan's tickets become visible and durable without changing the pack.

Works today

Shipped glue, with real examples.

Every card below is code that exists in this repo right now, not a promise. Each names the pack it bridges and how to see it work.

Works today superpowers → CCC board Subagents surface on your board. When a superpowers skill fans out Task subagents (dispatching-parallel-agents, subagent-driven-development), CCC counts them straight from the transcript and shows a subagent chip plus a live status rail on the parent session row. No change to superpowers. It just becomes visible.
Works today superpowers → watchtower → CCC A plan becomes a durable queue. The superpowers-to-watchtower skill lifts a superpowers plan out of its in-session scratch ledger: wt import turns it into board-visible Watchtower tickets, then optionally dispatches one CCC lane per ticket that closes with a summary. Session-local tracking becomes fleet-wide.
Works today CCC + gstack browse A lane that actually looks. The fleet-verify skill spawns one CCC lane that drives gstack browse (puppeteer as fallback) against the running app, verifies the exact change you describe, and reports a visual verdict with a screenshot. The check a diff read and a unit test cannot make.
Works today GET /api/skills CCC knows your packs. A read-only endpoint inventories CCC's own bundled skills plus every installed third-party pack, each annotated with honest flags: does it spawn subagents, is it fleet-aware, does it drive a browser. mtime-cached and stdlib-only, so a dashboard can poll it for free.
Works today Total Recall Cross-session search, already wired. CCC queries Total Recall for the sidebar session search and offers a launcher for its dashboard when the pack is installed. The memory a sibling session wrote is one search away.
Works today Watchtower (wt) The queue engine underneath. CCC imports watchtower.queue as its primary queue engine and can route inject and ask through wt send and wt ask. The board and the CLI read the same ground truth.

One flow, end to end

Plan in superpowers, run it as a fleet.

The superpowers-to-watchtower bridge, start to finish. Preview first, apply, then dispatch. Nothing here is pseudocode.

# 1. Superpowers wrote a plan (writing-plans / brainstorming output)
#    Preview the ticket split. `wt import` is preview by default: your dry run.
wt import my-feature-plan.md -q MYFEATURE

# 2. Looks right? Apply it. The plan is now durable, board-visible tickets.
wt import my-feature-plan.md -q MYFEATURE --apply --type feature
wt status -q MYFEATURE --json     # depth + oldest-open age, not "stuck"

# 3. Optional: one CCC lane per ticket. Each claims, does, closes with a summary.
curl -s -X POST "$CCC_URL/api/sessions/spawn" -d '{
  "prompt": "Read ticket MYFEATURE-3, claim it, do only that work,
             verify, then wt close it with a truthful summary.",
  "repo_path": "/abs/repo", "report_to": "<your-session-id>" }'

Falls back honestly: no wt means superpowers' own ledger; CCC down means in-process Task subagents. Never pretends a lane ran.

At a glance

The pack matrix.

How each installed pack relates to a CCC fleet, and whether the integration is wired today. Populated live by /api/skills.

Pack	What it owns	Fleet seam	Status
Superpowers	Plan, TDD, subagent fan-out, review	Subagents surface on the board; plan imports to a queue	wired
gstack browse	Headless browser QA and screenshots	A spawned lane drives it for visual verification	wired
Watchtower	Durable ticket queues and workers	CCC's primary queue engine and message backend	wired
Total Recall	Cross-session persistent memory	Sidebar search today; systematized fleet memory next	partly wired
Token Optimizer	Context and token cost auditing	fleet-auditor into CCC's health strip	roadmap
PostHog	Analytics, flags, error tracking	Not wired; listed for honesty	roadmap

As of 2026-07-17. Presence is detected live; the fleet-synergy flags are a curated, honest annotation. A pack being installed does not mean it is wired, and the matrix says which is which.

Roadmap, not done

What we have not built yet.

Naming these plainly is the point. They are real gaps, not shipped features described in the future tense.

Roadmap Total Recall Systematized fleet memory. Today CCC searches Total Recall. Not yet: every lane writing its outcome tagged by queue and ticket, so siblings recall what was already tried without being told. The plumbing exists; the convention does not.
Roadmap Token Optimizer Cost in the health strip. Token Optimizer's fleet-auditor already reads across agent systems. Surfacing its cross-session cost numbers inside CCC's own health strip is not built.
Roadmap superpowers subagent-driven-development Plan progress on the board. The chip shows how many subagents are running. It does not yet read the .superpowers/sdd/progress.md ledger to show which plan task is done and at which commit range.
Roadmap PostHog and the long tail The rest of the 256 packs. Most installed packs are detected and listed, nothing more. Wiring is deliberate and case by case, not a blanket claim of compatibility.