Kilroy

Kilroy is a local-first "software factory" CLI that turns English requirements into an Attractor pipeline (a directed graph in Graphviz DOT), then executes that pipeline node-by-node with tool-using coding agents in an isolated git worktree.

What Are Attractors?

An Attractor is a DOT-based pipeline runner for multi-stage AI workflows:

• A pipeline is a digraph written in Graphviz DOT syntax.
• Nodes represent stages (LLM/coding-agent tasks, human gates, tool steps, conditionals, parallel fan-out, etc.).
• Edges represent control flow, including conditions and retry loops.
• The engine checkpoints progress and (in Kilroy) commits after each node so runs can be resumed and audited.

Spec references:

• docs/strongdm/attractor/attractor-spec.md (graph DSL + engine semantics)
• docs/strongdm/attractor/coding-agent-loop-spec.md (tool-using coding agent loop)
• docs/strongdm/attractor/unified-llm-spec.md (provider-neutral LLM client)
• docs/strongdm/attractor/kilroy-metaspec.md (this repo's pinned decisions)

StrongDM Links

The Attractor + CXDB specs vendored into docs/strongdm/ originate from StrongDM:

• StrongDM: https://strongdm.com/
• CXDB: https://github.com/strongdm/cxdb
• In-repo specs: docs/strongdm/attractor/

CLI Quickstart

go build -o kilroy ./cmd/kilroy

# Turn requirements into a pipeline graph
./kilroy attractor ingest -o pipeline.dot "Build a Go CLI link checker with robots.txt support"

# Validate DOT structure/syntax
./kilroy attractor validate --graph pipeline.dot

# Execute the pipeline (requires a run config; see metaspec)
./kilroy attractor run --graph pipeline.dot --config run.yaml

For the run.yaml schema, CXDB requirements, and execution model, read docs/strongdm/attractor/kilroy-metaspec.md.

Skills: "Use The Skill To Use This Repo"

This repo ships Codex/Claude-style skill documents under skills/:

• skills/using-kilroy/SKILL.md: how to operate the kilroy attractor workflow (ingest/validate/run/resume).
• skills/english-to-dotfile/SKILL.md: how to turn requirements into a valid .dot pipeline (used by kilroy attractor ingest).

To use them with a coding agent that supports SKILL.md documents:

1. Add the skill(s) to your agent's skill search path (for example, copy or symlink skills/using-kilroy/ into your agent's skills directory).
2. In your prompt, tell the agent to use the using-kilroy skill when you want it to run Kilroy commands (and english-to-dotfile when you want a pipeline).

If you're using Kilroy's ingestor directly, it will auto-detect skills/english-to-dotfile/SKILL.md unless you pass --skill.