Bridle Examples

Reference implementations of the four key Bridle artifacts. Drop them into your project as a starting point.

File	Layer	Purpose
module.yaml	L0 module identity → L1 entry	Per-module manifest with identity, contracts, lifecycle, signals
intent.yaml	L2 intent expressible	Desired-state declaration with 4 blocks: business / contract / quality / lifecycle
reconciler.py	L3 autonomous loop	Reference reconciliation loop with R0–R5 gradient routing
CLAUDE.md	L3 Harness · system context	Drop-in CLAUDE.md / cursor-rules / copilot-instructions for AI agents

Reading order

1. Read module.yaml — understand what L1 state visibility looks like in YAML
2. Read intent.yaml — see how a business goal becomes a verifiable contract
3. Read CLAUDE.md — the AI agent’s bounded operating instructions
4. Read reconciler.py — how all of the above flow into one continuous loop

Adapting to your project

These are illustrations, not production-ready code. Adapt to your stack:

• Reconciler runtime: Temporal / Dapr / GitHub Actions cron / K8s Controller-runtime — see three-leaps-bootstrap.md §6.1 (P6.3) for stack-specific options
• Manifest schema validation: ajv / gojsonschema / jsonschema / NJsonSchema (CI must enforce)
• Audit store: append-only markdown (starter) → SQLite → EventStoreDB / Postgres (scale)
• Signal collection: each upstream tool has its own API (Codecov / archtest / sonar / OTel collector / GitHub API)

See ../three-leaps.md for the full methodology and ../three-leaps-bootstrap.md for the 35-capability bootstrap path that builds the infrastructure these examples assume.