SIX OPERATORS.
ZERO BABYSITTING.
SHIP THE TARGET.
CT6 is a six-strong tactical unit of long-lived AI operators. Hand them a target — trivial or fortified — and they recon, breach, hold, and exfiltrate with a clean commit. Multi-agent doctrine under the hood. A relentless fireteam in practice. Hard or easy. They take the objective.
Not a copilot. A fireteam that holds the objective.
AO mapped before contact
Cartographer + integration explorers recon the codebase, route graph, and external surfaces before a single line is touched.
Named operators, own context
Long-lived teammates — each with its own 1M context and callsign — fight on a shared task board. Lead calls the shots.
Rules of engagement, enforced
Hook-enforced review gates. 12-field evidence packet. Independent reviewer ≠ teammate. No green, no exfil.
Auto-spawned fix teams
Every contact becomes a Solution Requirement; a new squad rolls until the originating check goes green.
10-Phase Strike Plan
100%-coverage hard gate. 12-condition exit checklist.
- 01openspec validate --all --strict must be valid.
- 02Every requirement has ≥ 1 measurable scenario.
- 03Reuse decisions cite real files in CODEBASE_MAP.
- 04No duplicate capabilities. Phase 2 cannot start until clean.
Phenotypes — deployable architectures, reused by default.
A library of pre-made, generalized, deployable architectures — blueprint + parameterized scaffold (starter code + OpenTofu) + metadata — under phenotypes/. Three seeds ship: user-management, config-management, ai-management. Seed a run, or let the pipeline propose a reuse-first match — surfaced for confirmation, never applied silently.
--phenotype user-management
--label my-arch
Authorize the squad.
Take the objective.
Open-source under MIT. Built for senior operators who'd rather brief a unit than babysit a copilot. Brief. Authorize. Exfil with a clean commit.