rordi-ai

rordi-ai/loopbreaker

Plugin Claude Code

Bounded agent delivery from shaping through review: five skills and a local MCP behavior contract.

rordi-ai/loopbreaker

Plugin Claude Code

Shape, score, plan, implement, and review software against a local behavior contract with bounded review and explicit shipping decisions.

rordi-ai/loopbreaker

MCP server Claude CodeCodexCursor +2

MCP server "loopbreaker" as configured in rordi-ai/loopbreaker. Runs ./mcp/server.bundle.mjs with node.

rordi-ai/loopbreaker

Instructions file CodexOpenCode

Instructions for rordi-ai/loopbreaker: Loopbreaker is an agent-facing CLI and local MCP server. Keep machine output compact and stable. CLI stdout is TOON; diagnostics belong on stderr.

rordi-ai/loopbreaker

Instructions file

Instructions for rordi-ai/loopbreaker, a project described as: Local-first MCP review graph extracted from Rordi: bounded agent review, attributable evidence, and explicit shipping decisions.

rordi-ai/loopbreaker

Agent

Bounded implementation worker for exactly one named work unit of a frozen Loopbreaker behavior contract. Spawn one per work unit with isolation "worktree", passing the issue ID, the work unit, its enforced behaviors verbatim, and the proof plan. It implements and tests inside its worktree and reports back; it never…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Interview the founder until every required shape field traces to a human answer, then record the discovery record for approval. Use before shaping any new issue — shape cannot reach proceed without an approved discovery record. Use when a request arrives as an idea, when a shape field has no human behind it, or when…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Implement a feature from a planning-ready, frozen Loopbreaker behavior contract. Use when coding should begin, continue, or be repaired against named behaviors and work units, with minimal scope, production wiring, repository-native verification, and evidence recorded without allowing implementation to redefine…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Orchestrate a full Loopbreaker delivery as a three-role meta-agent system - a root orchestrator holding all state authority, isolated per-work-unit implementation workers, and an independent cross-vendor CLI reviewer that records its own findings over MCP. Use when a feature spans multiple work units, when…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Turn a shaped product slice into a frozen, executable feature contract and healthy planning profile backed by Loopbreaker. Use when scope is agreed and a builder needs behavior children, work-unit traceability, proportionate proof plans, production wiring, rollback, deterministic planning health, and an imported…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Perform planning-gated, bounded invariant and contract-driven code review using the Loopbreaker substrate. Use for a first review, repair re-review, planning-health preflight, review-finding validation, production-wiring checks, ship decisions, or any review at risk of scope expansion, repeated fixes, pass-four…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Independently review a shaped feature and its planning contract before implementation using Loopbreaker's persisted two-plus-one planning-review gate. Use when shape and planning appear ready, implementation admission must be decided, planning findings need re-review, or planning review risks becoming subjective…

rordi-ai/loopbreaker

Skill Claude CodeCodex

Shape a product idea, initiative, or epic into a bounded strategic bet before feature planning. Use when a request is still an idea, the outcome or appetite is unclear, multiple slices compete, scope needs cutting, reversibility matters, or a builder needs to decide shape, spike, park, or proceed.