GitHub - liuhaotian2024-prog/ystar-bridge-labs: A real company operated by AI agents, governed by Y*gov

3 min read Original article ↗

Y* Bridge Labs is the CEO and company behavior center for the Y* agent-company runtime. It is where the company mission, owner decisions, business artifacts, mission-command reports, external-validation planning, and CEO runtime sessions live.

This repository is not the canonical governance engine and it is not the provider executor. It produces and routes company decisions; Y-star-gov governs them; gov-mcp carries them to tool/provider boundaries in dry-run or approved execution modes.

Current role in the system

The current baseline comes from the full-repository E87R baseline and the E88 runtime binding work. The division of responsibilities is:

Repository Canonical role
ystar-bridge-labs CEO/company behavior center, mission command, business reasoning, owner decisions, reports, pre-action packets, post-action residuals.
Y-star-gov Governance reflex center, deterministic validators, runtime hooks, CIEUStore formal governance records.
gov-mcp Provider/tool execution boundary, outbound policy, dry-run/no-send receipts, future owner-activated execution preflight.
K9Audit Separate stronger hash-chain evidence ledger. It is not currently a write path from this repo.

What is implemented here

CEO behavior center

The existing CEO-facing behavior center lives under:

  • office/aiden_meeting_room/aiden_response_engine.py
  • office/aiden_meeting_room/company_context_loader.py
  • office/aiden_meeting_room/aiden_intent_classifier.py
  • office/aiden_meeting_room/meeting_memory.py

It reads repository-grounded company context, classifies owner intent, returns a CEO-level response, and keeps external side effects out of the response path.

Mission command and runtime bridge

The mission-command layer lives under office/mission_command/. It now includes:

  • e84_ystar_gov_ceo_cognitive_os_call_adapter.py - direct Y-star-gov validator call adapter.
  • e85_ceo_cognitive_os_runtime_bridge.py - CEO major-action runtime envelope and decision routing.
  • e87_ceo_runtime_session.py - end-to-end CEO runtime session runner.

The E87/E88 runtime session runner proves this internal chain:

bridge-labs behavior center
-> CEO pre-action packet
-> Y-star-gov validate_and_write_ceo_runtime_envelope(...)
-> formal Y-star-gov CIEUStore pre-action record
-> bridge-labs decision route
-> gov-mcp dry-run receipt when the action is provider/tool-bound
-> post-action residual
-> formal Y-star-gov CIEUStore residual record
-> next-action recommendation

Baseline reports

The canonical baseline reports are under:

  • operations/baseline/e87r_full_repo_baseline/

Important files include:

  • baseline_summary.json
  • architecture_evidence_map.json
  • stable_vocabulary_and_owner_map.json
  • l5_truth_table.json
  • final_goal_gap_analysis.json
  • next_engineering_roadmap.json
  • current_runtime_status_after_e88_bind_existing_labs_behavior_center_to_runtime_cieu_loop.json

Current L5 truth table

Do not describe the project as simply "L5 complete." Use this taxonomy:

Layer Current status Meaning
L5-A Runtime Foundation Complete for internal runtime foundation One internal CEO session can pass through behavior center, Y-star-gov runtime governance, formal CIEUStore writes, gov-mcp dry-run, and post-action residual closure.
L5-B CEO Intelligence Loop Partial The repo has intelligence artifacts and reports, but still needs a runtime/compiler flow for full repo recall, route generation, counterfactual comparison, commercial sharpness, and learning updates before every major action.
L5-C Controlled External Action Partial dry-run only gov-mcp dry-run/no-send boundary exists. Live provider execution is not enabled by this milestone.
L5-D Revenue/Customer/Payment Loop Absent or not executed No customer validation, paid signal, pricing validation, payment loop, or revenue loop is claimed.

What this repository does not claim

This repository does not claim:

  • live external outreach;
  • customer validation;
  • paid signal;
  • pricing validation;
  • compliance or legal proof;
  • production deployment;
  • live provider execution;
  • K9Audit ledger integration;
  • complete L5 revenue/customer/payment loop.

Key validation commands

Useful targeted checks:

python3 -m py_compile office/mission_command/e87_ceo_runtime_session.py
pytest -q tests/office/test_e87_ceo_runtime_session.py
pytest -q tests/office/test_e85_ceo_cognitive_os_runtime_bridge.py
pytest -q tests/office/test_e87r_full_repo_baseline_reports.py

Delivery model

Changes to this repository are delivered through the host-local repository delivery bridge when normal push from a sandbox is unavailable:

  • scripts/repository_delivery_bridge_submit.py
  • scripts/repository_delivery_bridge_worker.py
  • scripts/repository_delivery_bridge_status.py

The delivery bridge applies an allowlisted payload, runs targeted validation, commits, pushes, and records the remote-confirmed hash.

Next engineering direction

The next major engineering loop should upgrade L5-B: turn the CEO intelligence loop into a real runtime/compiler path that performs repository and history recall, candidate route generation, counterfactual comparison, commercial sharpness, adversarial critique, pre-action CIEU prediction, and next-action selection before every major action.