Skip to content

pursue-sdk-quality.md

Latest commit

 

History

History
50 lines (39 loc) · 2.23 KB

pursue-sdk-quality.md

File metadata and controls

50 lines (39 loc) · 2.23 KB

Pursuit: SDK Quality — Principal Eng Standard

Generation 1 — SHIPPED (PR #1361)

  • Security criticals: deterministic RNG, constant-time auth, key zeroing, job dispatch panic
  • Lint discipline: removed 5 dangerous workspace-wide allows
  • Panic elimination: keystore, eigenlayer registration, runner config
  • Result: 5.5 → 6.5/10

Generation 2 — ALL REMAINING HIGH/MEDIUM/STRATEGIC

Date: 2026-04-04 Status: building

Thesis

Seal every crack, harden every surface. Gen 1 stopped the bleeding. Gen 2 closes every remaining HIGH and MEDIUM finding, fixes the structural bloat, and adds the missing production hardening. Target: 9/10 across all categories.

Track A — Soundness & Unsafe

  1. Add // SAFETY: comments to unsafe impl Send/Sync in runner + manager, or eliminate them
  2. Fix unsafe env var mutation in remote-providers (use Mutex or scoped approach)
  3. Fix TOCTOU on key file permissions (create with restricted permissions from the start)
  4. Fix transmute endianness in JobId (use from_le_bytes/to_le_bytes)

Track B — Network Hardening

  1. Add per-peer rate limiting to P2P protocol (request + gossip)
  2. Add size-bounded deserialization for bincode on network messages
  3. Replace SSH command denylist with allowlist pattern

Track C — Async & Concurrency

  1. Replace QoS busy-wait polling with tokio::sync::watch channel
  2. Add backpressure to producer/consumer pipeline (bounded channels)

Track D — Clippy Deep Clean

  1. Audit remaining ~100 suppressed clippy lints — remove unjustified, fix warnings
  2. Target: only genuinely stylistic lints remain suppressed

Track E — TODO/Dead Code Purge

  1. Audit all TODO/FIXME in production code — resolve meaningful ones, delete stale ones
  2. Remove dead code surfaced by lint changes
  3. Clean up hardcoded values (service_id=1, blueprint_id=1 in QoS)

Track F — Build & Structure

  1. Populate workspace-hack with hakari output for compile-time dedup
  2. Audit and document the meta-crate strategy (keep/merge decision with rationale)

Strategic (deferred to Gen 3 — require protocol design decisions)

  • Key rotation for operators
  • Pre-submission slashing protection
  • MEV-protected result submission
  • cargo tangle dev zero-config mode