ward fuzz <id> - Ward

Usage

ward fuzz A1
ward fuzz u2   # ids are case-insensitive — normalized to uppercase internally
# or, from source:
npm run ward -- fuzz A1

What it does

Runs a single invariant end to end, in debug log level, and prints:

The invariant’s id, name, source citation, and category.
A PASS/FAIL badge and duration.
Its evidenceSummary — the same one-line summary ward test’s table shows.
The full evidence object, pretty-printed as JSON — every request sent and response received for every sub-case, unredacted.

This is the fastest way to see exactly what payload Ward built, exactly what mutation was applied, and exactly how the facilitator responded — the detail ward test’s compact table intentionally leaves out.

Valid invariant ids

U1, U2, U3, U4, U5, A1, A2, A3 — see the full Invariants Overview.

Example output

$ ward fuzz A1

Fuzzing A1 — atomic group integrity
Algorand atomic transaction groups (protocol primitive) + @x402/avm facilitator's
verifyFeePayerTransaction; USENIX Security '26 study — Free Shopping, Asset Theft,
Service Denial, Gas Abuse
category: algorand

Result: PASS (2143ms)
fee-payer leg amount > 0 (drains value instead of a pure self-payment): OK;
fee-payer leg receiver redirected away from self: OK;
fee-payer leg closes remainder balance to a third party: OK;
fee-payer leg rekeys the facilitator's own account: OK;
fee-payer leg fee inflated far past the reasonable cap: OK

--- evidence ---
{
  "baselineTransaction": "ABCD1234...",
  "subResults": [ ... ],
  "evidence": [
    { "label": "GET /supported (discovering fee-payer address)", "request": null, "response": { ... } },
    { "label": "baseline sponsored-fee settle (sanity check)", "request": { ... }, "response": { ... } },
    { "label": "verify — fee-payer leg amount > 0 ...", "request": { "original": { ... }, "tampered": { ... } }, "response": { ... } }
  ]
}

Exit code

Exits non-zero if the invariant failed — useful for scripting a one-off check without running the whole suite, e.g. in a bisecting workflow while debugging a facilitator change.

Re-render a full past run