asupersync-mega-skill
Made by JeffBuild, migrate, debug, and maintain Asupersync. Use when working with Tokio migration, Cx/Scope/cancellation, local tasks, Lab replay, browser/Wasm, protocols, databases, OTLP, or repository proof.
Works With
Quality and Utility come from canonical skill metadata. Community ratings appear separately as stars when subscriber reviews exist.
Subscribe to download
Public previews are available, but downloads, CLI sync, and premium content require an active subscription.
Sign In to Subscribe- Full skill content with deterministic versioning
- CLI sync + local search index
- Access to Jeffrey’s premium library
Save to your account
Saved skills live in your cloud list. Run jsm sync on each machine to pull them locally.
What's New with this Skill
4Updates, enhancements, and release notes for this skill.
Updated: asupersync-mega-skill, telling "on main" apart from "in the published crate"
Asupersync's workspace version reads 0.4.9, which does not tell you whether a given API is in the published crate. Registered unary gRPC serving over native H2 works on main at commit 3c73a334c and is absent from published v0.4.9. Release status is recorded per capability, and the routing matrix carries required and prohibited actions for each case, including five negative controls for when to avoid selecting the skill.
- The routing matrix has six columns (ID, Select, Prompt, First reference, Must do, Must not do); five negative controls mark prompts where the skill should not be selected at all
- Registered unary gRPC serving over native H2 works on main at commit 3c73a334c and is absent from published v0.4.9
- OTLP configuration and owned metrics, traces and logs (A2-A5) shipped in v0.4.9 behind the native metrics feature; A6-A11 remain open
- CompiledSupervisor computes deterministic restart plans, but a compiled tree is not restarted automatically today; live restart is per-actor via Scope::spawn_supervised_actor
- Actor and GenServer receive and drain loops yield after every eight continuously ready messages, which is a bounded scheduling point rather than a latency guarantee
- Established execute and query keep their SqliteError signatures; separately named *_diagnosed methods return the additive SqliteOperationError
- RQ SSH secret delivery ships at commit 515d96e7f with bounded protected stdin and the key environment removed from bootstrap subprocesses; --rq-auth-key-stdin is preferred over the legacy --rq-auth-key-hex
- The JS/TS packages are workspace-local at 0.4.9 and unpublished on npm, with scoped GA for main-thread and dedicated-worker consumers while the public Rust browser API stays a preview
The routing matrix in SELF-TEST.md has six columns: ID, Select, Prompt, First reference, Must do, and Must not do. Twelve positive cases name the reference to open first and the action to take; five negative controls carry Select=no, marking prompts where this skill should not be chosen. local-task has to enter the owning runtime's real worker-local lane and must not treat run_test as a local executor. web-handler has to use the matching AsyncCxFnHandler arity and must not pass a raw async function to get or post. migration-compat has to require exact bridge compile and runti
asupersync-mega-skill v5: the #[main] on-ramp, the v0.4.4 cancellation contract, and a nocrypto ATP board that beats tuned rsync
The August refresh moves the skill onto Asupersync's blessed on-ramp (#[asupersync::main], with Runtime::request_cx_with_budget as the ambient-free production Cx entry), states the v0.4.4 cancellation contract precisely, and rewrites the ATP evidence snapshot from a single banked cell to the MATRIX-235 ledger: a board-level nocrypto win over tuned rsync, a fully measured encrypted matrix, and a refuted-levers list so nobody re-chases dead optimizations.
- #[asupersync::main] is the blessed level-0 on-ramp, with #[test] and #[lab_test] as matching entry attributes and docs/onramp.md as the graduated path.
- Runtime::request_cx_with_budget(budget) is the production ambient-free Cx entry; Cx::for_request() and Cx::for_testing() stay test-harness material.
- The v0.4.4 cancellation contract, stated precisely: a typed result survives cancellation acknowledgement on ordinary Cx::spawn*, pre-first-poll cancellation keeps v0.4.3 semantics, and neither abort-always-wins nor value-always-survives is correct.
- race! and blocking select! are drain-correct: every loser is protocol-cancelled and drained before return, with the else and timeout: exceptions spelled out.
- Dependency sovereignty is teachable now: first-party base64/hex, an owned regex engine, protobuf wire plus OTLP, UTC time, and minimal DER, with per-module cutover status.
- Repo rules now in the skill: a v0.4.3 backwards-compat release gate with more than 12 production consumers, the v0.3.10 patch-bump trap on record, and a release-blocking parked-task cancellation lane on a date-pinned nightly.
- ATP evidence graduated from one banked cell to a ledger: a board-level nocrypto win over tuned rsync from a 56-row sweep, a fully measured 25/25 encrypted matrix, and receiver RSS down from 882 MB to 12 MB.
- A refuted-levers list (MATRIX-222 through 229 and 234) keeps future agents from re-chasing flow-control credit, BBR startup shapes, window raises, and other dead ends.
Bootstrap teaching moved to the graduated on-ramp. #[asupersync::main] is the blessed level-0 entry, with #[asupersync::test] and #[asupersync::lab_test] as the matching test attributes and docs/onramp.md as the four-level path; the explicit RuntimeBuilder route remains for apps that need to own the runtime object. Runtime::request_cx_with_budget(budget) is now the blessed ambient-free production Cx entry, and Cx::for_request() / Cx::for_testing() stay demoted to test-harness material.
asupersync-mega-skill catches up with the July API: RuntimeBuilder bootstrap, Evidence-First Operator Cards, and honest ATP benchmark accounting
Asupersync's public surface moved in July and the skill moved with it. The production bootstrap seam is now RuntimeBuilder plus block_on with a runtime-owned Cx; Cx::for_request() is test-harness material. Five Evidence-First Operator Cards teach agents to classify every repo-internal claim (banked, parity, candidate, stale, and four more), gate ATP benchmark talk on real matrix evidence, and route serious migrations through the repo's read-only readiness planner.
- Production bootstrap, corrected for July: RuntimeBuilder::current_thread().build() plus block_on with a runtime-owned Cx; Cx::for_request() and Cx::for_testing() are test/internal-harness paths, not production architecture.
- Admission-sensitive spawning: try_spawn / try_spawn_with_cx with explicit SpawnError handling; Cx::spawn / Cx::spawn_in for region-owned tasks, with Scope::spawn_registered scoped to boot/test callers.
- Claim gate: every repo-internal claim is classified banked, scoped-cell-only, parity, correct-but-slower, candidate, stale, blocked, or failed, and the no-claim boundary is stated in the same breath.
- ATP gate: benchmark wins require tuned rsync, release atp, crypto-symmetric setup, rate caps, and current matrix/ledger evidence; sha_ok, cargo check, or one favorable stale cell proves nothing.
- Current ATP snapshot: one banked RQ/nocrypto cell (500M/broken, 564.77s vs rsync 574.46s), a measured encrypted-QUIC 50M/good win, and MATRIX-210/211 named as implementation landings, not wins.
- Serious migrations start with the repo's read-only migration readiness planner (verdict, proof commands, phase plan); the skill's audit-target.sh is demoted to quick inventory.
- Generated API anchors from api_surface_map_v1.json: http::Client fluent builders, lab::ScenarioRunner and CancellationInjector, and web::Router middleware layers.
- NETWORKING-PROTOCOL-STACK and REPO-CONTRIBUTOR-GUIDE roughly doubled, SOURCE-MAP grew by about 60%; RaptorQ distributed guidance expanded; new portable scripts/validate.sh validates the skill itself.
Asupersync's API drifted, and a skill that teaches yesterday's bootstrap pattern produces code that is wrong today. This refresh realigns the whole teaching seam: RuntimeBuilder::current_thread().build() plus block_on installs an ambient runtime context, runtime-spawned tasks carry a runtime-owned Cx, and app code receives &Cx. Cx::for_request() and Cx::for_testing() are explicitly demoted to test and internal-harness paths. For admission-sensitive bootstrap the skill now routes through try_spawn / try_spawn_with_cx with explicit SpawnError handling, and it separates Cx::spawn / Cx::spawn_in (
New: Asupersync Mega Skill, replace Tokio with a cancel-correct async runtime
A guide to replacing Tokio-based Rust stacks with Asupersync, a spec-first, cancel-correct, capability-secure async runtime. Covers migration from tokio/axum/hyper/tonic and designing native Cx/region-based services, with 35 reference files.
- Step-by-step migration from Tokio, Axum, Hyper, and Tonic, with a compatibility boundary that keeps Tokio types out of business logic
- Native Cx/region-based service design patterns, choosing a greenfield or brownfield lane before touching code
- Cancel-correct async: structured concurrency and obligation tracking, so an unresolved permit, ack, or lease at region close is reported as an ObligationLeak
- Debugging Asupersync internals with the lab runtime, DPOR, and replay traces