feat: implement solver reputation score, fix slash ordering, add bond…

[solarix-x]

Title:

feat: solver reputation score, slash ordering fix, bond-conservation
proptest, solver-registry design

─────────────────────────────────────────────────────────────────────

Description:

Closes #29, Closes #43, Closes #47, Closes #46

Summary

Four issues tackled in one pass — they share the same data model and
touching
lib.rs once is cleaner than four separate PRs with merge conflicts.

---

#29 — Fix slash_solver transfer-before-persist ordering

Problem: slash_solver was doing the token transfer to
fee_recipient
before writing the updated SolverRecord and IntentRecord to
storage. Since
the function is permissionless, a caller could (in theory) exploit
the gap
between the transfer and the state commit.

Fix: Both .set(...) calls now happen before
client.transfer(...). The
intent state is already Open in storage by the time any token
moves, so a
back-to-back call on the same intent_id hits the
IntentNotAccepted guard
immediately.

Test added: double_slash_second_call_rejected — calls
slash_solver
twice on the same intent and asserts the second returns
Err(IntentNotAccepted).

---

#47 — Reputation score calculation

New internal helper: compute_reputation_score(&SolverRecord) -> u32

Returns a score in 0–9 999 basis points derived from existing
fill history:

base = fills_completed / (fills_completed + fills_failed) [success
rate]
decay = VOLUME_SCALE / (VOLUME_SCALE + total_volume + 1) [volume
factor]
score = base × (1 − 0.1 × decay) × 10 000

• Zero attempts → 0. All failures → 0.
• Perfect solver with no volume → 9 000 bps (90%).
• Perfect solver with high volume → approaches 10 000 bps.
• All arithmetic is integer-only; no panics possible.

New view function: get_reputation_score(solver) -> Option<u32>
— returns
None for unregistered solvers.

Tests added: zero fills, all failures, perfect+no volume,
perfect+high
volume, partial failures vs perfect, unregistered solver, post-fill
score.

---

#43 — Bond-conservation invariant proptest

New dev-dependency: proptest = "1.5.0"

New file: src/proptest_bond.rs — a state-machine harness with 3
solvers
and up to 20 random steps per run (register, deregister, withdraw,
accept+slash).

After every step it asserts:

contract.bond_balance == Σ solver_record.bond_amount

CI: New dedicated proptest job runs PROPTEST_CASES=256 so it
doesn't
slow down the main contract job.

---

#46 — solver_registry design document

New file: docs/solver-registry-design.md

Covers:

• Reputation score formula (carries forward from Implement reputation score calculation from existing SolverRecord fill history #47)
• 5-tier table (Unranked → Platinum) with min bond, min score, and
  perks
• Contract boundary analysis — recommends Option A (registry owns
  solver
  state, intent_settlement calls in via cross-contract calls)
• Full cross-contract interface (eligibility check, record_fill,
  slash, stake,
  unstake, migrate)
• Fill-window extension and reduced slash % mechanics per tier
• One-time migration plan from existing intent_settlement storage
• Open questions (fee rebate %, tier downgrade timing, NFT badges,
  governance)

---

Testing

All existing tests are unchanged. New tests are additive. Proptest
harness is
gated behind --features testutils and runs in its own CI job.

Files changed

File	Change
intent_settlement/src/lib.rs	slash ordering fix,
compute_reputation_score, get_reputation_score
intent_settlement/src/test.rs	double-slash test, reputation
score tests
intent_settlement/src/proptest_bond.rs	new — bond-conservation
harness
intent_settlement/Cargo.toml	add proptest dev-dependency
.github/workflows/ci.yml	add proptest CI job
docs/solver-registry-design.md	new — solver registry design doc

|

[drips-wave]

@solarix-x Great news! 🎉 Based on an automated assessment of this PR, the linked Wave issue(s) no longer count against your application limits.

You can now already apply to more issues while waiting for a review of this PR. Keep up the great work! 🚀

Learn more about application limits