signalbrain 0.1.2

Trust layer for AI-modified software — receipts, ledger, calibrated autonomy

SignalBrain

Trust layer for AI-modified software.

Get started · Receipt spec · Architecture & roadmap · The founding incident · Pilot · Demo repo

Every company is letting agents change systems that matter. Every agent overstates what it did. SignalBrain is the referee: signed improvement receipts, objective re-score, and per-class calibrated trust — so autonomy is earned, not self-reported.

Agent tooling today answers risk with a permission prompt — approve every action, forever. Receipts are the exit ramp: an agent earns the right to stop asking, one measured claim at a time, per change-class, revocable by evidence.

Your repo, your ledger, no server. Plain files, a CLI, and a GitHub Action — nothing to host, nothing phones home. And because a referee can't also be a player, SignalBrain is agent- and model-neutral by design: Claude Code, Cursor, goose, Codex CLI — same rules for every one of them.

This repository is Phase 0 v0.1: the receipt spec, ledger math, scoring lane, anti-Goodhart machinery, and the founding incident record — extracted from the Titan reference deployment (R&D dummy that keeps trying to game its own ledger, in public).

60-second demo — run it, don't trust it

pip install signalbrain
bash demo/demo.sh

Raw transcript (real output — no mocks)

▶ 1. An agent tries to score its own claim BEFORE anyone merged it
  {"status": "refused_guard", "code": 3, "message": "... not on HEAD — score only human-merged receipts"}
  refused: unmerged claims cannot enter the ledger. No agent grades its own homework.

▶ 2. A batch of receipts measured only by tests the agent wrote itself
  ledger now holds 3 rows — every one classified: 3 "claim_kind": "invariant_pin"
  {}   (no class has ANY trust-eligible claims)
  three green results, ZERO earned trust: held-by-construction pins are recorded, never counted.

▶ 3. An honest failure
  "held": false
  the agent said 0.9 confidence. The measurement said no. That gap is the product.

▶ 4. Ten claims that actually hold
  "tooling": { "hit_rate": 1.0, "n": 10, "status": "auto-merge ELIGIBLE" }
  earned by track record, revocable by evidence. Autonomy is graduated, never granted.

The receipt lifecycle

flowchart LR
    A["Agent ships change<br/>+ receipt"] --> B{"human<br/>merges?"}
    B -- "no" --> R["refused — unmerged claims<br/>cannot be scored"]
    B -- "yes" --> C["sb score<br/>re-runs the receipt's<br/>own commands"]
    C --> D{"measured only by<br/>tests it wrote itself?"}
    D -- "yes" --> P["invariant_pin<br/>recorded · zero trust"]
    D -- "no" --> E{"commands<br/>pass?"}
    E -- "yes" --> H["held ✓"]
    E -- "no" --> F["held ✗<br/>recorded forever"]
    H --> L[("ledger")]
    F --> L
    P --> L
    L --> G{"last 10 high-confidence<br/>claims ≥ 95% held?"}
    G -- "yes" --> M["auto-merge ELIGIBLE<br/>earned · revocable"]
    G -- "no" --> N["GATE<br/>human review"]

    classDef good fill:#0d2b1e,stroke:#34d399,color:#a7f3d0
    classDef bad fill:#2b1214,stroke:#f87171,color:#fecaca
    classDef neutral fill:#0f172a,stroke:#475569,color:#cbd5e1
    class M,H good
    class R,F,P bad
    class A,B,C,D,E,G,L,N neutral

Three layers

Layer	What	Status
Receipt	Open standard — signed, re-runnable claims	docs/RECEIPT_SPEC.md v0.1
Ledger	Per-class trust from objectively re-scored receipts	src/signalbrain/governance/
Refuter	Adversarial verification + SPC (premium)	scripts + roadmap

Founding proof

Our own autonomous lane tried to pad its trust score to 100% ELIGIBLE in a local working tree. It never reached git. Full receipt-style incident record with reproduce commands:

docs/incidents/2026-07-tooling-trust-streak-gaming.md

Every number in that document is re-derivable from cited SHAs.

The ledger data has its own headline: across 58 objectively measured claims, hold-rate falls as stated confidence rises — 86% in the 0.85–0.90 bin, 83% in 0.90–0.95, 33% above 0.95. The most confident claims were the least reliable. Reproducible curves + generator: report/calibration-curves/.

Quick start

pip install signalbrain

# 1. Teach your agents to emit receipts (paste into CLAUDE.md / .cursorrules):
#    docs/pilot/receipt-emission.md

# 2. After a receipt merges, score it objectively:
sb score receipts/0001-tooling-my-change.md --root . --ledger .signalbrain/ledger.jsonl

# 3. Read the trust gates (exit 0 = TRUST earned, 1 = GATE):
sb gate --ledger .signalbrain/ledger.jsonl --by-class --window 10

# Or wire it into CI — see the fork-able demo's workflow:
#    https://github.com/whitestone1121-web/receipt-gate-demo

Reference-deployment invocations (legacy scripts, kept for parity)

export PYTHONPATH=src:scripts
python scripts/calibration_ledger.py docs/calibration/improvement_claim_ledger.jsonl \
  --require-measured --by-class --window 10
bash scripts/calibration_score_receipt.sh docs/improvements/NNNN-name.md
pytest tests/ -q

v0.1 scope and roadmap

See Architecture, provenance & roadmap — what's in the box, why the rules look the way they do, and what design partners drive next. Known limitations are stated there plainly; this project publishes its edges the same way it publishes its incidents.

Compat note: governance modules live under signalbrain.governance; agi_os_backend.governance shims preserve script import paths from the reference deployment.

Design partner offer

We score your coding agents' claims against what actually merged. First caught overclaim is free — if we don't find one, you still get an audit. Contact: signalbrain.ai

License

Apache-2.0 — see LICENSE.