Anti-Abuse & Probation

A monetizable reputation is a target. If a fake edge could become payout-eligible without scrutiny, the marketplace would fill with manipulation and collapse. These defenses are what make a rented edge trustworthy.

Probation

A new BYO source can't cash in on day one:

• It starts in probation — no payout eligibility until it has enough signal history and age to be judged fairly.
• This removes the incentive to spin up disposable Agents hunting for a lucky streak.
• Probation ends when the Agent meets the sample/age bar (see Scoring & Proven).

Attacks & defenses

Slashing

The $SGHI stake (see Seats & Staking) is the bond behind good behavior:

• Confirmed manipulation — self-dealing, wash trading, sustained over-rate spam — can slash part or all of the stake.
• Slashing is tied to evidence the engine can show (linkage graphs, volume composition), not to discretion alone.
• A slashed source loses its quality bond and its standing — reputation doesn't survive it.
• If a source detaches from an Agent NFT, its slashing/probation history follows the source operator record.

Review & governance

Slashing should not be a silent, one-click admin action. The intended flow is staged:

Detection signal
  ↓
Automatic flag + evidence bundle
  ↓
Temporary freeze if user harm is likely
  ↓
Manual review
  ↓
Creator response / appeal window
  ↓
Decision: clear, warning, probation extension, freeze, partial slash, or full slash
  ↓
Public outcome record

Who decides?

Early phase may use a protocol review panel. Longer term, high-severity slashes can move toward transparent governance or delegated risk councils.

Enforcement ladder

Slash amount should scale with severity, confidence, repeat history, and user harm. Not every flag becomes a slash.

Evidence & public record

The public outcome should include enough evidence to make the decision auditable without leaking private alpha or doxxing users.

Appeal window

Creators should have a defined response window before final slashing unless there is active harm:

• Standard appeal window — target 72 hours after notice.
• Emergency freeze — immediate freeze allowed, but slash still waits for review unless harm is already proven.
• Creator response — creator can submit logs, key-compromise evidence, infra outage evidence, or counter-analysis.
• Appeal outcome — decision can be upheld, reduced, reversed, or converted to warning/probation.
• Restoration — if reversed, frozen payouts and status should be restored where possible.

False-positive controls

• No slash from one weak signal. Slashing requires multiple evidence points or one high-confidence proof.
• Separate flagging from punishment. Automated systems flag; review decides.
• Reason codes required. Every enforcement action needs a machine-readable reason and human-readable explanation.
• Reproducible evidence bundle. Reviewers should be able to replay the scoring/abuse calculation from recorded inputs.
• Grace for disclosed key compromise. Fast self-reporting can reduce penalty if the creator stops the key and cooperates.
• Threshold tuning. Detection thresholds should be tested on historical benign sources before launch.

Open problem: renter leak

One hard case is still being worked out: a renter takes a 7-day rental, screenshots or re-broadcasts the signals, and resells them.

Directions under consideration:

• Alpha decay — the highest-value window is short, so leaked signals age out fast.
• Per-renter watermarking — subtle differences let a leak be traced to its source.
• Trace + revoke — identified leakers lose access.

← Back to BYO Overview.