Skip to content

fix(spec): add eth_sendRawTransaction to HYPERLIQUID as a write (MAG-4426) - #195

Open
nimrod-teich wants to merge 1 commit into
mainfrom
mag-4426
Open

nimrod-teich wants to merge 1 commit into
mainfrom
mag-4426

Conversation

@nimrod-teich

@nimrod-teich nimrod-teich commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

#Closes MAG-4426

Update spec: Hyperliquid (HYPERLIQUID)

Scoped to one method on request: eth_sendRawTransaction. The full update-mode re-research was not run. Nothing was removed.

Count
Methods added 1
Collections / extensions / directives added 0
Values corrected 0
Reported, not changed 1

Why

  • HyperEVM serves eth_sendRawTransaction, but HYPERLIQUID did not list it.
  • The smart-router relays an unlisted method under a default entry, Default-eth_sendRawTransaction. That entry has 20 compute units, stateful 0 and no timeout (defaultApiContainer in smart-router protocol/chainlib/base_chain_parser.go).
  • So every HyperEVM transaction went to one node and was retried on failure. It was never broadcast. A retry after a lost reply can report a landed transaction as failed.
  • relay_end (smart-router #543) records these relays as reads. The dashboard (MAG-4422) can find HyperEVM transactions only by method name.

Added

Target Definition Source
HYPERLIQUID jsonrpc POST eth_sendRawTransaction 10 CU, deterministic false, stateful 1, hanging_api, DEFAULT latest, timeout_ms 10000 Live probe and docs, below

HYPERLIQUIDT imports HYPERLIQUID, so the testnet inherits the method. The relay window becomes 12 s: timeout_ms 10000 plus twice the 1 s block time. As an unlisted method it was 2 s.

The definition matches the catalogue. All 11 own definitions are stateful 1 at 10 CU. Ten are hanging_api with DEFAULT latest. The most common explicit timeout_ms is 10000, used by CONFLUX, GNOSIS, LIT and QUAI.

Evidence. Both public RPCs serve the method. A malformed transaction gets -32602, and an unknown method gets -32601. Probed 2026-10-19:

$ curl -s -X POST https://rpc.hyperliquid.xyz/evm -H 'Content-Type: application/json' \
    --data '{"jsonrpc":"2.0","id":1,"method":"eth_sendRawTransaction","params":["0x00"]}'
{"jsonrpc":"2.0","id":1,"error":{"code":-32602,"message":"failed to decode signed transaction"}}
$ curl -s -X POST https://rpc.hyperliquid.xyz/evm -H 'Content-Type: application/json' \
    --data '{"jsonrpc":"2.0","id":1,"method":"eth_notAMethod","params":[]}'
{"jsonrpc":"2.0","id":1,"error":{"code":-32601,"message":"Method not found"}}

https://rpc.hyperliquid-testnet.xyz/evm answers the same. The HyperEVM page says wallets add the chain by RPC URL and chain ID, and that "Pre-EIP-155 transactions are accepted as with many EVM chains." The JSON-RPC page does not list the method.

Corrected

None.

Reported, not changed

  • eth_usingBigBlocks is on the JSON-RPC page's custom-endpoint list but not in the spec. It is out of scope here.

Guard output

check_update_diff.sh               RESULT: PASS (1 added, 0 modified — all declared; nothing removed)
check_hanging_api.sh --base        HYPERLIQUID/jsonrpc/eth_sendRawTransaction|hanging ok (cu=10, timeout_ms=10000ms >= 1000ms implied)
check_stateful.sh --base           HYPERLIQUID/jsonrpc/eth_sendRawTransaction|write, stateful=1
check_method_schema.sh             HYPERLIQUID/jsonrpc/eth_sendRawTransaction|schema ok   (no FAIL rows)
check_unused_fields.sh             RESULT: PASS (no removed fields)
check_disabled_count.sh --expect 0 RESULT: PASS (file and claim agree on 0 disabled)
check_node_admin_rpcs.sh           OK: hyperliquid.json serves no new node-operator control from a base collection.
test_*.sh guard self-tests         19 of 19 pass

Test plan

  • The guards and self-tests above.
  • Booted a local smart-router on HYPERLIQUID and HYPERLIQUIDT and sent eth_sendRawTransaction through it. Both nets broadcast it as a write to both upstreams; main's spec relays it as Default-eth_sendRawTransaction, a read, to one. Results.

The auto-started Spec Pipeline run is cancelled, because this PR adds one method to a chain that is already onboarded. A /rerun-<phase> comment restarts it.

…4426)

HyperEVM serves eth_sendRawTransaction, but HYPERLIQUID did not list it.
The smart-router relays an unlisted method under a default entry with
stateful 0, so every HyperEVM transaction went to one node, was retried
on failure, and was never broadcast.

The new entry matches the catalogue's other definitions: 10 compute
units, deterministic false, stateful 1, hanging_api, DEFAULT latest and
timeout_ms 10000. HYPERLIQUIDT imports HYPERLIQUID, so the testnet gets
it too.
@magmadevs-bot

Copy link
Copy Markdown
Contributor

🤖 Spec pipeline started on PR open — running Phase 8 → 11. Results post here as comments. View run

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Spec pipeline failed (started at Phase 8). See the run log.

Retry by commenting:
/rerun-probe (optionally /rerun-probe mainnet=https://your-node/rpc`, or mainnet=use=PAID_RPC_1 for a stored secret node)`

To also re-run every later phase on top of the fix, use /rerun-from 8 instead.

@nimrod-teich

Copy link
Copy Markdown
Contributor Author

Manual smart-router run: eth_sendRawTransaction on HYPERLIQUID and HYPERLIQUIDT (2026-10-19)

Result: PASS. On the branch spec, eth_sendRawTransaction binds to the spec entry on both nets and is relayed as a write to every upstream. On main's spec, the same request is relayed as Default-eth_sendRawTransaction, a read, to one upstream. The reads are unaffected.

The router ran the static spec catalog (--use-static-spec), once with main's hyperliquid.json and once with this branch's (99b0cfe). Every request went through the router. The only write payload was the literal "0x00", which no node accepts as a transaction, so no transaction reached any chain.

The binary is smart-router PR #543 at 151cfa9a, built with go build ./cmd/smartrouter. It is main plus the relay_end log event, which writes one line per relay. It does not change how the router picks upstreams, and it was used only for that line.

Upstreams used

Two upstreams per net, so the stateful fan-out is visible.

Net Spec (chain id) Upstream name URL
mainnet HYPERLIQUID (0x3e7) hl-main-official https://rpc.hyperliquid.xyz/evm
mainnet HYPERLIQUID (0x3e7) hl-main-drpc https://hyperliquid.drpc.org
testnet HYPERLIQUIDT (0x3e6) hl-test-official https://rpc.hyperliquid-testnet.xyz/evm
testnet HYPERLIQUIDT (0x3e6) hl-test-drpc https://hyperliquid-testnet.drpc.org

Second upstreams were vetted with smartrouter health (through the router, no direct node traffic). Eight candidates were checked and all passed chain-id: mainnet dRPC, HyperLend, Stakely, thirdweb 999, Hypurrscan, Purroof; testnet dRPC, thirdweb 998. dRPC was picked for both nets. The two dRPC hostnames share one edge, so they count as one origin for the polling budget.

Boot verifications

Branch spec, both nets (all four chain-id checks pass, both trackers failed=0):

[+] verified successfully chainId=HYPERLIQUID  nodeUrl=https://rpc.hyperliquid.xyz/[redacted]          verification=chain-id rawData=0x3e7
[+] verified successfully chainId=HYPERLIQUID  nodeUrl=https://hyperliquid.drpc.org                     verification=chain-id rawData=0x3e7
[+] verified successfully chainId=HYPERLIQUIDT nodeUrl=https://rpc.hyperliquid-testnet.xyz/[redacted]  verification=chain-id rawData=0x3e6
[+] verified successfully chainId=HYPERLIQUIDT nodeUrl=https://hyperliquid-testnet.drpc.org             verification=chain-id rawData=0x3e6
ChainTracker initialization complete chainID=HYPERLIQUID  success=2 failed=0
ChainTracker initialization complete chainID=HYPERLIQUIDT success=2 failed=0

Main spec, mainnet only: the same two HYPERLIQUID lines (rawData=0x3e7) and ChainTracker initialization complete chainID=HYPERLIQUID success=2 failed=0. The same lines repeat in every boot of this run: 2 branch-spec boots and 2 main-spec boots, none failed.

Results

Surface Result Evidence (via router)
HYPERLIQUID boot, branch spec PASS chain-id 0x3e7 verified on both upstreams, tracker success=2 failed=0
HYPERLIQUIDT boot, branch spec PASS chain-id 0x3e6 verified on both upstreams, tracker success=2 failed=0
eth_sendRawTransaction on HYPERLIQUID PASS relay_end names eth_sendRawTransaction (no Default- prefix), stateful: true, selection: stateful, 2 attempts in one batch, one on each upstream
eth_sendRawTransaction on HYPERLIQUIDT (inherited by import) PASS same shape as mainnet: bound to the spec api, stateful, 2 attempts in one batch
eth_chainId, eth_blockNumber read controls PASS stateless, answered on the first attempt, correct values on both nets

eth_sendRawTransaction before and after

Payload: {"jsonrpc":"2.0","id":1,"method":"eth_sendRawTransaction","params":["0x00"]}.

Spec Net Method as relay_end names it stateful selection Attempts (count, upstreams) Outcome Client response
main mainnet HYPERLIQUID Default-eth_sendRawTransaction false stateless 1: hl-main-drpc. Not retried: the node error is non-retryable. node_error, stop_reason NonRetryableNodeError HTTP 200, error -32602 "failed to decode signed transaction"
branch (mag-4426) mainnet HYPERLIQUID eth_sendRawTransaction true stateful 2: hl-main-official (node_error) and hl-main-drpc (node_error), both batch 0 node_error, stop_reason Stateful HTTP 200, error -32602 "failed to decode signed transaction"
branch (mag-4426) testnet HYPERLIQUIDT eth_sendRawTransaction true stateful 2: hl-test-official (node_error) and hl-test-drpc (HTTP 500 transport_error, retryable), both batch 0 node_error, stop_reason Stateful HTTP 200, error -32602 "failed to decode signed transaction"

Samples: main's spec was booted twice (once with text logs, once with JSON logs) and gave the same result both times. Branch mainnet gave the same shape in 2 of 2 boots. Branch testnet gave the row above in 2 of 3 probes. The first testnet probe hit a transient failure, described in the run notes.

The router metrics agree. On the main-spec boot the series carries function="Default-eth_sendRawTransaction" (spec=HYPERLIQUID). On the branch boots it carries function="eth_sendRawTransaction" for both spec=HYPERLIQUID and spec=HYPERLIQUIDT, and no Default- label exists.

Read controls (branch spec, boot A)

Net Method Client response served_by selection attempts
mainnet eth_chainId 0x3e7, HTTP 200 hl-main-official stateless none listed (first attempt succeeded)
testnet eth_chainId 0x3e6, HTTP 200 hl-test-drpc stateless none listed
mainnet eth_blockNumber 0x2eb0e86, HTTP 200 hl-main-official stateless none listed
testnet eth_blockNumber 0x4033cd5, HTTP 200 hl-test-drpc stateless none listed

relay_end evidence (trimmed)

Fields kept: GUID, method, stateful, chain, selection, outcome, stop_reason, served_by, status, error_code, latency_ms, and per attempt the upstream, outcome, error_code, batch and returned_to_client. The message is kept only on transport errors.

# BEFORE: main spec, mainnet (JSON-log boot)
{"GUID":"17454565850620385586","method":"Default-eth_sendRawTransaction","stateful":false,"chain":"HYPERLIQUID","selection":"stateless","outcome":"node_error","stop_reason":"NonRetryableNodeError","served_by":"hl-main-drpc","status":200,"error_code":"USER_INVALID_PARAMS","latency_ms":347,"attempts":[{"upstream":"hl-main-drpc","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":true}]}
# AFTER: branch spec, mainnet eth_sendRawTransaction (boot A, then boot B)
{"GUID":"14905865349366843811","method":"eth_sendRawTransaction","stateful":true,"chain":"HYPERLIQUID","selection":"stateful","outcome":"node_error","stop_reason":"Stateful","served_by":"hl-main-drpc","status":200,"error_code":"USER_INVALID_PARAMS","latency_ms":511,"attempts":[{"upstream":"hl-main-official","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":false},{"upstream":"hl-main-drpc","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":true}]}
{"GUID":"5271671241563742117","method":"eth_sendRawTransaction","stateful":true,"chain":"HYPERLIQUID","selection":"stateful","outcome":"node_error","stop_reason":"Stateful","served_by":"hl-main-official","status":200,"error_code":"USER_INVALID_PARAMS","latency_ms":449,"attempts":[{"upstream":"hl-main-official","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":true},{"upstream":"hl-main-drpc","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":false}]}
# AFTER: branch spec, testnet eth_sendRawTransaction (boot B, then the repeat in boot B)
{"GUID":"1222173839442634986","method":"eth_sendRawTransaction","stateful":true,"chain":"HYPERLIQUIDT","selection":"stateful","outcome":"node_error","stop_reason":"Stateful","served_by":"hl-test-official","status":200,"error_code":"USER_INVALID_PARAMS","latency_ms":648,"attempts":[{"upstream":"hl-test-official","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":true},{"upstream":"hl-test-drpc","outcome":"transport_error","error_code":"NODE_INTERNAL_ERROR","batch":0,"returned_to_client":false,"message":"HTTP 500 500: Internal node error"}]}
{"GUID":"12345776258411497741","method":"eth_sendRawTransaction","stateful":true,"chain":"HYPERLIQUIDT","selection":"stateful","outcome":"node_error","stop_reason":"Stateful","served_by":"hl-test-official","status":200,"error_code":"USER_INVALID_PARAMS","latency_ms":634,"attempts":[{"upstream":"hl-test-official","outcome":"node_error","error_code":"USER_INVALID_PARAMS","batch":0,"returned_to_client":true},{"upstream":"hl-test-drpc","outcome":"transport_error","error_code":"NODE_INTERNAL_ERROR","batch":0,"returned_to_client":false,"message":"HTTP 500 500: Internal node error"}]}
# AFTER: branch spec, testnet eth_sendRawTransaction (boot A, transient: no upstream answered)
{"GUID":"11247256384956806628","method":"eth_sendRawTransaction","stateful":true,"chain":"HYPERLIQUIDT","selection":"stateful","outcome":"no_answer","stop_reason":"Stateful","served_by":"","status":0,"error_code":"PROTOCOL_NETWORK_UNREACHABLE","latency_ms":3958,"attempts":[{"upstream":"hl-test-drpc","outcome":"transport_error","error_code":"NODE_INTERNAL_ERROR","batch":0,"returned_to_client":false,"message":"HTTP 500 500: Internal node error"},{"upstream":"hl-test-official","outcome":"transport_error","error_code":"PROTOCOL_NETWORK_UNREACHABLE","batch":0,"returned_to_client":false,"message":"http request failed: Post \"https://rpc.hyperliquid-testnet.xyz/[redacted]\":"}]}
# read controls (boot A): eth_chainId mainnet, eth_chainId testnet, eth_blockNumber mainnet, eth_blockNumber testnet
{"GUID":"10024698239820848055","method":"eth_chainId","stateful":false,"chain":"HYPERLIQUID","selection":"stateless","outcome":"ok","stop_reason":"Success","served_by":"hl-main-official","status":200,"error_code":"","latency_ms":278,"attempts":[]}
{"GUID":"4007538043606472888","method":"eth_chainId","stateful":false,"chain":"HYPERLIQUIDT","selection":"stateless","outcome":"ok","stop_reason":"Success","served_by":"hl-test-drpc","status":200,"error_code":"","latency_ms":86,"attempts":[]}
{"GUID":"16279853132845558812","method":"eth_blockNumber","stateful":false,"chain":"HYPERLIQUID","selection":"stateless","outcome":"ok","stop_reason":"Success","served_by":"hl-main-official","status":200,"error_code":"","latency_ms":276,"attempts":[]}
{"GUID":"4623112277571697298","method":"eth_blockNumber","stateful":false,"chain":"HYPERLIQUIDT","selection":"stateless","outcome":"ok","stop_reason":"Success","served_by":"hl-test-drpc","status":200,"error_code":"","latency_ms":90,"attempts":[]}

Run notes

  • The "retried" path is not shown on main's spec. The node error is not retryable, so the read stopped after one attempt on one upstream. The run does show that main's spec never broadcasts.
  • The first testnet write in boot A got no answer. The official testnet host failed at dial time with a local "network is unreachable" error, and dRPC testnet answered HTTP 500. The two later testnet writes and every read in that boot were clean. It is a network blip, not a spec effect.
  • dRPC testnet answers the "0x00" payload with HTTP 500 in 3 of 3 probes. dRPC mainnet and both official hosts return -32602. With the broadcast, the client still gets the official host's node error.
  • Flags: --chain-tracker-poll-divisor 0.25, so each upstream is polled every 4 s and the public RPCs' rate limits are not hit. --min-relay-timeout 5s, so a cold TLS connection cannot add a timeout to the evidence. The relay retry limit stays at its default of 2. No 429 in any boot.
  • Neither spec has a SUBSCRIBE directive, across the import closure, so no WebSocket leg was run.
  • 7 client requests and 12 upstream sends in total. No valid signed transaction was built or sent. Every router process was stopped afterwards.

@magmadevs-bot

Copy link
Copy Markdown
Contributor

🤖 Spec pipeline started on PR open — running Phase 8 → 11. Results post here as comments. View run

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Phase 8 — smart-router boot + method probe

PARSE: OK · VERIFY: OK · TESTNET_VERIFY: OK · ADDONS: 1 tested-ok / 0 failed / 0 not-testable · PASS=27 FAIL=0 SKIP=1 UNPROBED=0 WARN=0 TIMEOUT=0 LOG_WARN=1 · BLOCK_TIME: mainnet=980ms testnet=980ms

PR change under test — eth_sendRawTransaction binding: PASS on both networks. relay_end names the method eth_sendRawTransaction (not Default-), stateful=true, selection="stateful"; client got -32602 "failed to decode signed transaction" from the deliberately-undecodable 0x00 payload (the correct outcome — nothing was signed or broadcast).


Method Probe Report — hyperliquid

Generated: 2026-10-20T15:07:27Z
Router image: ghcr.io/magma-devs/smart-router:main
Router config: /tmp/sr_hyperliquid.yml (mainnet), /tmp/sr_hyperliquid_testnet.yml (testnet)
Spec variant: HYPERLIQUID (jsonrpc) / HYPERLIQUIDT (jsonrpc)
Upstreams probed: https://rpc.hyperliquid.xyz/evm (mainnet, SINGLE_UPSTREAM), https://rpc.hyperliquid-testnet.xyz/evm (testnet, SINGLE_UPSTREAM)

Only one upstream URL per network was provided, so one direct-rpc block per network was configured. Multi-node rotation and WARN-DISAGREEMENT detection were NOT exercised — this run cannot and does not claim multi-node coverage.

Transport availability

The spec declares zero SUBSCRIBE parse directives in the resolved import closure (HYPERLIQUID.imports=[], HYPERLIQUIDT.imports=["HYPERLIQUID"]), so no ws leg is required to boot and both listeners were booted plain-http-only per the Step 1b hard-requirement rule (N/A here).

Per this run's transport-probing requirement, both node URLs were additionally probed over ws:

URL ws result
wss://rpc.hyperliquid.xyz/evm REFUSED — HTTP/1.1 405 Method Not Allowed with allow: POST returned by the CloudFront/nginx front door on a raw TLS websocket-upgrade handshake (confirmed via a direct TLS socket + manual Upgrade: websocket request, since curl's libcurl build here lacks wss:// support)
wss://rpc.hyperliquid-testnet.xyz/evm REFUSED — identical 405 Method Not Allowed / allow: POST

Since the spec has no subscription methods, both refusals are expected and NOT a defect or a boot blocker.

Parse-directive & verification runtime check (Step 3.5) — mainnet (HYPERLIQUID)

Note: the :main image ships with block-hash polling (fork detection, which is what actually exercises GET_BLOCK_BY_NUM) off by default (--enable-fork-detection is required; confirmed via --help and the boot log line "block-hash polling (fork detection) resolved" enabled=false reason=off-operator-choice on an initial default boot). The router was rebooted with --enable-fork-detection to get authoritative evidence for GET_BLOCK_BY_NUM; this is a router CLI flag, not a spec or YAML-config change.

Check Verdict Evidence
GET_BLOCKNUM (router latest_block) OK smartrouter_latest_block=49056242 (metric), rpc_endpoint_fetch_latest_success=13, 0 fails
GET_BLOCK_BY_NUM (log hash-reads) OK 17–18 Chain Tracker Updated block hashes / Chain Tracker read a block Hash lines across two measurement windows, with --enable-fork-detection; fetch_block_* metric is the known-dead metric (never emitted)
Parse-signature log lines none no failed to parse response / ParseBlockHashFromReplyAndDecode / etc. signatures in any boot or probe window
Verification Verdict Providers OK/failed Notes
chain-id OK 1/1 expected_value=0x3e7 matched live eth_chainId=0x3e7; confirmed on the archive-addon-free boot used for Steps 3.5/4

Operational note (not a spec defect): on a separate boot made solely to probe the archive extension (with addons: [archive] added to the sole node-urls entry per Step 1c-iii), the chain-id verification stopped being logged at all ("no verifications for url" fired only for the archive-tagged internal route; the base/"" route produced no visible validation line, though the chain tracker and relays continued to function normally). This looks like a router node-url/addon-batching quirk when a single URL is tagged with an addon and is unrelated to the hyperliquid spec — the chain-id VERIFY:OK verdict above comes from the untagged boot used for the authoritative Steps 3.5/4 run, so it is not downgraded by this observation. Flagging it as a router-behavior note worth a maintainer's attention, not a hyperliquid spec fix.

Addon & extension coverage (Step 1c + probes)

The spec declares ONE extension, archive (cu_multiplier: 5, rule.block: 127), and ZERO collection_data.add_on addons. No verification in the spec is gated on archive (values[].extension is absent from the sole chain-id verification), so per Step 1c-ii this was tested without a gating verification: a direct historical-block call (500 blocks behind tip) to the upstream, which returned full block data (not pruned).

A known structural gap carried over from the orchestrator's pre-resolved facts: there is no pruning verification on this spec despite the archive extension being declared, so GET_EARLIEST_BLOCK (eth_getBlockByNumber("earliest")) is never invoked by the router. Confirmed independently: a direct eth_getBlockByNumber(["earliest", false]) call to the mainnet upstream returns {"error":{"code":-32603,"message":"invalid block height: 0"}} — the node does not serve true genesis — so even if a pruning verification existed with the stock expected_value convention it would need adjusting for this chain. This is a pre-existing spec gap to flag for Phase 9, not something introduced or fixed here.

Name Type Upstreams supporting Standalone Boot verification Router probe Classification
archive extension 1/1 (mainnet upstream) — n/a (no verification gates archive) PASS — with addons: [archive] configured, router logged supportingProviderCount:1, no "extension unavailable" warning, and returned a valid historical block TESTED_OK

Evidence detail: a first archive probe attempt (before adding addons: [archive] to the config) returned a valid result but logged a WARN — "a caller's lava-extension names an extension no node offers, serving without it" — because the upstream hadn't been tagged yet. After adding addons: [archive] to the node-urls entry (Step 1c-iii) and rebooting, the same call routed cleanly with supportingProviderCount:1 and no warning. Classified TESTED_OK on the corrected config.

GET_EARLIEST_BLOCK — NOT_EXERCISED (nothing in the spec calls it; no pruning verification exists to invoke it), consistent with the pre-resolved facts for this run.

Method probe table (27 stateless methods, through router at localhost:3360)

All 27 non-stateful methods were probed unpaced; zero HTTP 429s were encountered against rpc.hyperliquid.xyz, so spaced mode was never entered.

Method Classification Upstream notes Notes
net_version PASS — result="999"
web3_clientVersion PASS — result="hyperliquid evm Mainnet"
eth_blockNumber PASS —
eth_call PASS — to=live contract addr, data=0x
eth_chainId PASS — 0x3e7
eth_estimateGas PASS —
eth_feeHistory PASS —
eth_gasPrice PASS —
eth_getBalance PASS — live address
eth_getBlockByHash PASS — live block hash
eth_getBlockByNumber PASS — live block number
eth_getBlockReceipts PASS — live block number
eth_getBlockTransactionCountByHash PASS —
eth_getBlockTransactionCountByNumber PASS —
eth_getCode PASS — live contract address, returned bytecode
eth_getLogs PASS — live block range
eth_getStorageAt PASS —
eth_getTransactionByBlockHashAndIndex PASS — live block hash + index 0
eth_getTransactionByBlockNumberAndIndex PASS —
eth_getTransactionByHash PASS — live tx hash
eth_getTransactionCount PASS —
eth_getTransactionReceipt PASS — live tx hash
eth_maxPriorityFeePerGas PASS —
eth_syncing PASS — result=false
eth_bigBlockGasPrice PASS — Hyperliquid-specific method, resolved fine
eth_getSystemTxsByBlockHash PASS — Hyperliquid-specific method, resolved fine
eth_getSystemTxsByBlockNumber PASS — Hyperliquid-specific method, resolved fine
eth_sendRawTransaction SKIP (normal loop) / see "PR change under test" below — category.stateful: 1 — never broadcast; binding verified separately below

All parameterized methods used live values resolved from the upstream immediately before probing (latest block number/hash, a tx hash + sender/receiver address drawn from that block, tx index 0) — no hardcoded placeholders were used, so the re-probe-once gate was never triggered (no method needed it).

Log-scan findings (probe window, Step 4.5)

Level Associated method error excerpt
WARN archive-extension probe (not a listed method) "a caller's lava-extension names an extension no node offers, serving without it" — fired on the first archive probe attempt before addons: [archive] was added to the config; resolved (no recurrence) after the Step 1c-iii config fix, see Addon coverage section above

No other non-benign warn/error/fatal/panic lines appeared during either the mainnet or testnet probe windows. No fatal/panic lines at any point.

PR change under test — eth_sendRawTransaction binding (PR #195)

Per Step 4, eth_sendRawTransaction has category.stateful: 1 and was SKIPped by the normal probe loop (never broadcast). Binding was instead verified by sending the deliberately-undecodable payload {"jsonrpc":"2.0","id":1,"method":"eth_sendRawTransaction","params":["0x00"]} through the router, which cannot decode into a signed transaction and therefore cannot move funds or be broadcast.

Mainnet (HYPERLIQUID, localhost:3360): client response {"error":{"code":-32602,"message":"failed to decode signed transaction"}}. relay_end log: method="eth_sendRawTransaction" (correctly bound, not Default-eth_sendRawTransaction), stateful=true, selection="stateful", error_code="USER_INVALID_PARAMS", served_by="hyperliquid-upstream-1".

Testnet (HYPERLIQUIDT, localhost:3460, during Step 7): same payload, same result on the first attempt (no re-probe needed — no transient 500/no_answer was observed this run): {"error":{"code":-32602,"message":"failed to decode signed transaction"}}, relay_end log: method="eth_sendRawTransaction", stateful=true, selection="stateful", error_code="USER_INVALID_PARAMS", served_by="hyperliquid-testnet-upstream-1".

Both results exactly match the expected outcome from the prior manual run on this branch. The -32602 is the correct, desired outcome of an intentionally-undecodable payload, not a failure.

Verdict: PASS-existence / PASS-binding on both mainnet and testnet — eth_sendRawTransaction is correctly bound to its spec entry (not falling back to Default-), correctly classified stateful/selection: stateful, and correctly routed to the configured upstream, with zero risk of any transaction being broadcast.

Testnet verification pass (Step 7)

TESTNET_VERIFY: OK

Verification Verdict Providers OK/failed Notes
chain-id OK 1/1 expected_value=0x3e6 matched live eth_chainId=0x3e6 on rpc.hyperliquid-testnet.xyz/evm

PARSE_BLOCKNUM: OK (smartrouter_latest_block=67418286). PARSE_BLOCK_BY_NUM: OK (8 Chain Tracker Updated block hashes/read-hash log lines with --enable-fork-detection). No parse-signature lines. Boot was clean (no panic/fatal/exclusion lines).

Empirical block time (Step 8)

Network RPC Empirical (ms) Spec effective (ms) Drift Verdict
mainnet https://rpc.hyperliquid.xyz/evm 980 1000 (HYPERLIQUID average_block_time) -2% OK
testnet https://rpc.hyperliquid-testnet.xyz/evm 980 1000 (HYPERLIQUIDT declares its own average_block_time: 1000, same value) -2% OK

Both measurements are well within the 20% threshold; no BLOCK_TIME_MISMATCH.

Teardown

Both sr_hyperliquid and sr_hyperliquid_testnet containers removed (docker rm -f); docker ps -a --filter name=sr_hyperliquid confirms no leftover containers.

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Phase 9 — parallel reviewers (3 fresh contexts)

reviewer 1  TALLY: CRITICAL=0 MEDIUM=3 MINOR=1
reviewer 2  TALLY: CRITICAL=0 MEDIUM=2 MINOR=2
reviewer 3  TALLY: CRITICAL=0 MEDIUM=2 MINOR=2

Three distinct reports (distinct md5s), every cited evidence line within hyperliquid.json (561 lines) — no stale-state reviewer. Zero CRITICAL findings from all three. All three independently reviewed the PR's changed surface and passed it.

Verdict on the PR's changed surface — eth_sendRawTransaction

All three reviewers passed it, agreeing with the Phase 8 probe: category.stateful: 1 is correct for a broadcast method, category.hanging_api: true with timeout_ms: 10000 satisfies the max(1s, compute_units × 100ms) floor (10 CU × 100ms = 1s), compute_units: 10 is in band for an EVM write, and block_parsing DEFAULT/latest is the right shape for a method that takes no block argument. HYPERLIQUIDT picks it up through imports with no entry of its own, which is correct.

Merged gap list (deduplicated; reviewer agreement in brackets)

# Severity Finding On main? Agreement
1 MEDIUM archive extension on the mainnet collection has no pruning verification, so GET_EARLIEST_BLOCK is never executed (Phase 8 confirmed it NOT_EXERCISED). Reviewers 2 and 3 add that the stock expected_value convention would not hold here anyway — mainnet's earliest call returns -32603 invalid block height: 0. pre-existing 3/3
2 MEDIUM / MINOR allowed_block_lag_for_qos_sync: 5 is tighter than the formula value and than peer-chain convention for a 1000 ms block time. pre-existing 3/3 (R1 MEDIUM, R2+R3 MINOR)
3 MEDIUM eth_getSystemTxsByBlockHash / eth_getSystemTxsByBlockNumber use block_parsing EMPTY despite taking a block hash/number — inconsistent with this same spec's convention for the other hash/number-keyed queries. pre-existing 2/3 (R1, R3)
4 MEDIUM Standard EVM filter-polling methods (eth_newFilter family) are absent from the spec although the chain exposes them on the same /evm path. pre-existing 1/3 (R2)
5 MINOR eth_gasPrice carries an unusually tight timeout_ms: 1000. pre-existing 1/3 (R2)
6 MINOR No API-docs path was supplied to the reviewers, so a full method-coverage diff was not possible — a review-coverage limitation, not a spec defect. n/a 2/3 (R1, R3)

Every finding is pre-existing on main — none was introduced by this PR. Rows 1–5 go to the Phase 10 consolidation; row 6 is dropped as a coverage note.

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Phase 10 — synthesize gaps + fix pass

Consolidation: 0 critical, 4 medium; 0 disable-suggestions stripped; 0 in-scope, 4 pre-existing. MINOR gaps dropped per the phase rules. Full list on the runner at docs/hyperliquid/FIX_LIST.md; reproduced below.

Auto-decision at the fix gate: no spec edit applied (conservative default, recorded)

This is an unattended run, so the fix gate was auto-decided with the conservative default. The consolidation verified mechanically — git show origin/main:hyperliquid.json diffed against the working tree — that the only difference between main and this branch is the added eth_sendRawTransaction block. Every surviving gap is therefore byte-identical to main: this PR introduced none of them, and nothing on the surface this PR actually changes needs fixing.

Applying them anyway would push four unrelated edits to a production mainnet+testnet spec under a PR whose stated scope is one write method — the exact silent-drift failure the skill guards against — and three of the four cannot be applied safely on the evidence this run has:

# Severity Gap Blind-apply risk Decision
1 MEDIUM archive extension has no pruning verification, so GET_EARLIEST_BLOCK is never executed (Phase 8: NOT_EXERCISED). Raised by all 3 reviewers. HIGH — mainnet's eth_getBlockByNumber(["earliest",false]) returns -32603 invalid block height: 0, so the stock expected_value: "0x0" convention can never be satisfied. Adding it would fail provider startup verification and exclude every upstream at boot. Not applied. Needs chain-specific calibration of the earliest-serviceable block, not a template copy.
2 MEDIUM eth_getSystemTxsByBlockHash / …ByBlockNumber use block_parsing: EMPTY despite block-keyed names. Raised by reviewers 1, 3. MEDIUM — inferred from naming only. No API docs were available to this run; reviewer 3 labelled it PLAUSIBLE, not CONFIRMED. Both methods PASS at the live tip today. Not applied. Needs the real parameter list from docs or node source first.
3 MEDIUM EVM filter-polling family (eth_newFilter etc.) absent from the spec. Raised by reviewer 2 (reviewer 3 rated the same fact MINOR and doubted it). LOW (pure addition) but unconfirmed, and adding 6–9 methods is well outside this PR's scope. Not applied. Belongs in its own update-mode PR with a docs-backed method list.
4 MEDIUM allowed_block_lag_for_qos_sync: 5 on both entries, vs 10 from the formula (10000 / 1000 ms) and the catalog convention for 1000 ms chains. Raised by all 3 (reviewer 1 MEDIUM; 2 and 3 MINOR, citing arbitrum.json as precedent for halving). LOW — provider-scoring tolerance only; gates nothing. Not applied. Lowest-risk of the four and the best candidate for a quick follow-up, but it still changes a live mainnet and testnet parameter outside this PR's scope, and two of three reviewers rated it MINOR with existing precedent.

Rows 1–4 are carried to the watch-list below rather than silently left: they are real, they are on main, and they want their own PRs.

Post-fix static gates (Phase 10a) — all PASS

check_method_schema   PASS (28/28 apis, incl. eth_sendRawTransaction — FAIL section empty)
check_unused_fields   PASS (no removed fields)
check_hanging_api     PASS  HYPERLIQUID/jsonrpc/eth_sendRawTransaction|hanging ok
                            (cu=10, timeout_ms=10000ms >= 1000ms implied)
check_stateful        PASS  eth_sendRawTransaction|write, stateful=1
                            eth_call|read, stateful=0 · eth_estimateGas|read, stateful=0

check_disabled_count.sh: 0 distinct disabled methods, 0 rows — unchanged from main. The Disabled-API Justifications ledger is therefore empty (none — no disabled APIs), and no <!-- disabled-count: N --> marker needs revising.

Removed-field governance guard: the spec carries none of the 15 fields smart-router dropped in #218 — no spec-level governance fields, no proposal.title/description, no top-level deposit, no extra_compute_units, no category.local, no category.subscription. Envelope is the canonical { "proposal": { "specs": [ … ] } }; compute_units is the only compute-unit input; the spec declares zero FUNCTION_TAG_SUBSCRIBE directives, so no method is a subscription.

Watch-list (needs a follow-up PR, not this one)

  1. Calibrate and add the pruning verification for the archive extension — or deliberately document why this chain cannot carry one. Until then GET_EARLIEST_BLOCK is dead weight in the spec.
  2. Confirm the eth_getSystemTxsByBlock* parameter list against HyperEVM docs/node source and set block_parsing accordingly.
  3. Confirm whether /evm serves the filter-polling family; if so, add it in update mode.
  4. Reconcile allowed_block_lag_for_qos_sync: 5 with the formula value 10 across both entries, or record the chain-specific reason for the deviation.

Working tree: hyperliquid.json is unmodified by this phase (git status clean) — there is no fix-pass edit for the pipeline's commit step to pick up.

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Phase 10b — smoke regression re-probe

SMOKE: OK — no regressions. Re-booted the dockerized smart-router against the on-disk spec and re-probed the deterministic set on both networks. (The spec is byte-identical to what Phase 8 probed, since Phase 10 applied zero fixes — the re-boot was run anyway, as the phase requires.)

PARSE: OK   ·  VERIFY: OK  ·  ADDONS: 1 tested-ok / 0 failed / 0 not-testable  ·  TESTNET_VERIFY: OK
Probe Phase 8 Phase 10b Result
GET_BLOCKNUM (eth_blockNumber) PASS PASS 0x2ec8da2
chain-id verification (eth_chainId) PASS PASS 0x3e7
eth_bigBlockGasPrice PASS PASS 0x5f5e100
eth_blockNumber (sampled #1) PASS PASS as above
eth_call (sampled #3) PASS PASS 0x
eth_chainId (sampled #4) PASS PASS as above
eth_estimateGas (sampled #5) PASS PASS 0x5208

Sampled reads = the first five non-stateful APIs alphabetically. eth_sendRawTransaction is category.stateful: 1, so it is excluded from the sample and never broadcast.

  • PARSE — smartrouter_latest_block = 49057181 (mainnet) / 67419227 (testnet), fetch_latest_success nonzero, 0 fails, no parse-signature lines. GET_BLOCK_BY_NUM produced 0 hash-read lines on a default boot and 54 (mainnet) / 55 (testnet) after re-booting with --enable-fork-detection — the same router-flag observation Phase 8 made, reproduced on the :latest tag. This is a router default, not a spec defect.
  • VERIFY — chain-id 1/1 on both networks: mainnet 0x3e7, testnet 0x3e6, both matching live eth_chainId.
  • ADDONS — archive re-probed through the router with the Lava-Extension: archive header against a block 500 behind tip: supportingProviderCount:1, no extension-unavailable warning, full unpruned historical block returned. Matches Phase 8's TESTED_OK.
  • TESTNET_VERIFY — clean boot of HYPERLIQUIDT on a separate container, chain-id 1/1 0x3e6 live-matched, PARSE OK on both directives. The testnet upstream's previously-observed flakiness did not recur this run.
  • Log scan (probe window) — one non-benign error line, from an extra eth_sendRawTransaction binding re-check with the undecodable ["0x00"] payload: the expected -32602 failed to decode signed transaction, identical to Phase 8. No other warn/error lines; no fatal/panic at any point.
  • Teardown — both sr_hyperliquid and sr_hyperliquid_testnet removed; docker ps -a confirms no leftovers. No spec file edited.

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Phase 11 — final reviewer verdict

⚠️ CHANGES REQUESTED

TALLY: CRITICAL=0 MEDIUM=4 MINOR=1

The verdict is reported mechanically and honestly: the phase gate is CRITICAL=0 MEDIUM=0 → APPROVED, and 4 MEDIUM gaps remain, so this is CHANGES REQUESTED. It has not been upgraded.

What the verdict does and does not mean. The final reviewer verified mechanically (git show origin/main:hyperliquid.json diffed with jq -S .) that the only difference between main and this branch is the added eth_sendRawTransaction block — every other byte, including all four flagged fields, is byte-identical to main. So all 5 findings are PRE-EXISTING technical debt; zero were introduced by this PR, and the reviewer's own readiness line reads: "PR #195 is ready to merge as scoped." The CHANGES-REQUESTED label is the gate firing on whole-file MEDIUM count, not on anything this PR did.

The PR's change passed cleanly

eth_sendRawTransaction drew no findings. The reviewer confirmed it matches ethereum.json's (ETH1) configuration field-for-field on block_parsing, category, and compute_units, with timeout_ms: 10000 a reasonable chain-specific addition. Binding verified PASS on both mainnet and testnet in Phases 8 and 10b, with zero broadcast risk. HYPERLIQUIDT inherits it automatically through imports and correctly defines no apis of its own.

Remaining gaps — all pre-existing on main

# Severity Finding Fix risk
1 MEDIUM archive extension has no pruning verification; GET_EARLIEST_BLOCK is dead code. Mainnet's earliest call errors -32603 invalid block height: 0, so the stock expected_value: "0x0" template would exclude every provider at boot if copied blind. HIGH
2 MEDIUM eth_getSystemTxsByBlockHash / …ByBlockNumber use block_parsing: EMPTY despite block-keyed names. PLAUSIBLE, not CONFIRMED — no API docs available to settle the parameter list. MEDIUM
3 MEDIUM EVM filter-polling family absent from the spec with no rationale. LOW (pure addition)
4 MEDIUM allowed_block_lag_for_qos_sync: 5 on both entries vs the formula's 10 (10000 / 1000 ms); peers at the same block time (ink, katana, megaeth) use 10. LOW
5 MINOR No real API-docs path was available, so a doc-driven completeness diff could not be run — findings 2 and 3 are inference-based. Review-coverage limitation, not a spec defect. n/a

Findings explicitly checked and dismissed this pass

  • block_distance_for_finalized_data: 0 — standard for BFT/instant-finality chains (~40 specs in the catalog); consistent with HyperBFT. Not a defect.
  • blocks_in_finalization_proof: 1 — correct per the finality-typed rule (fast/instant finality → 1).
  • eth_gasPrice timeout_ms: 1000 — raised as MINOR in Phase 9, dropped here: ETH1's own eth_gasPrice carries the identical value, so it matches upstream convention.
  • Disabled APIs/collections — zero enabled: false apis or collections anywhere in the file, matching the empty Disabled-API Justifications ledger exactly. No CRITICAL finding on the disable gate.
  • Removed-field guard — RESULT: PASS (no removed fields).

Governance-field compliance (checked explicitly)

The spec carries none of the 15 fields smart-router removed in #218: no spec-level governance fields (min_stake_provider, providers_types, contributor, contributor_percentage, shares, identity, block_last_updated, reliability_threshold, data_reliability_enabled), no proposal.title / proposal.description, no top-level deposit, and no api-level extra_compute_units / category.local / category.subscription. Envelope is the canonical { "proposal": { "specs": [ … ] } }. compute_units is the only compute-unit input, and the spec declares zero FUNCTION_TAG_SUBSCRIBE directives, so no method is a subscription.

Recommended disposition

Merge PR #195 on its own merits, and open separate follow-ups for gaps 1–4. Gap 1 in particular must not be fixed by copying the standard pruning template — it needs this chain's real earliest-serviceable block determined first.

@magmadevs-bot

Copy link
Copy Markdown
Contributor

Phase 12 — closing summary checklist (tail of Phase 11; this pipeline run covered Phases 8–11)

#### File Validation
- ✓ JSON syntax valid                                    (jq clean at every phase gate; Phase 10a re-check)
- ~ All required fields present                          (Phase 6 did not run in this span; Phase 10a's
                                                          check_method_schema PASSed all 28 apis instead)
- ✓ No duplicate API names within any collection         (Phase 10a check_method_schema, 28/28 schema ok)
- ✓ Proper indentation/formatting                        (file untouched this run; diff vs main is the
                                                          single added api block and nothing else)

#### Configuration Verification
- ✓ Network parameters calculated per formulas           (~ one deviation found and reported, NOT fixed:
                                                          allowed_block_lag_for_qos_sync=5 vs formula 10 —
                                                          pre-existing on main, Phase 11 gap #4)
- ~ All APIs tested and working                          (Phase 8: PASS=27 FAIL=0 WARN=0 TIMEOUT=0
                                                          UNPROBED=0, SKIP=1 — eth_sendRawTransaction,
                                                          stateful, deliberately never broadcast.
                                                          Single upstream per network, so no multi-node
                                                          rotation or disagreement detection was possible)
- ~ Block parsing validated for each API                 (Phase 8 existence-tested; full per-API parse
                                                          validation needs production traffic)
- ✓ Parse directives executed by live router             (Phase 8 Step 3.5: PARSE=OK, VERIFY=OK;
                                                          Phase 10b re-check: PARSE=OK, VERIFY=OK.
                                                          GET_BLOCK_BY_NUM needed --enable-fork-detection
                                                          on both image tags — a router default, not a
                                                          spec defect. GET_EARLIEST_BLOCK is
                                                          NOT_EXERCISED: no pruning verification invokes
                                                          it — Phase 11 gap #1)
- ~ Addons & extensions tested                            (Phase 8 Step 1c: 1 tested-ok / 0 failed /
                                                          0 not-testable — the archive extension passed
                                                          through the router with the extension header and
                                                          re-passed in Phase 10b)
- ✓ Verifications pass on live nodes                     (chain-id live-matched on both networks in both
                                                          Phase 8 and Phase 10b: mainnet 0x3e7,
                                                          testnet 0x3e6)
- ☐ Compute units benchmarked under expected load        (user to measure)

#### Documentation (out of skill scope — manual reminder)
- ☐ SPEC_IMPLEMENTATION.md created
- ☐ API_REFERENCE.md created
- ☐ TESTING_GUIDE.md created
- ☐ QUICK_START.md created

#### Testnet vs Mainnet
- ✓ Mainnet spec complete                                (HYPERLIQUID, 28 apis, 0 disabled)
- ✓ Testnet spec inherits correctly                      (HYPERLIQUIDT imports HYPERLIQUID, defines no
                                                          apis of its own; Phase 11 confirmed it picks up
                                                          the new method automatically)
- ✓ Chain IDs verified for mainnet AND testnet           (live eth_chainId through the router on both,
                                                          twice — Phase 8 and Phase 10b)
- ✓ Both tested on respective networks                   (Phase 8 full probe + Step 7 testnet pass;
                                                          Phase 10b re-probe + testnet pass.
                                                          TESTNET_VERIFY: OK both times)

#### Governance Prep (out of skill scope — manual reminder)
- ☐ Proposal JSON formatted correctly                    (canonical { "proposal": { "specs": [ … ] } }
                                                          envelope confirmed; none of the 15 fields
                                                          removed in smart-router #218 are present)
- ☐ Community feedback gathered (if applicable)

Run outcome

Item Result
Phase 8 — boot + probe PARSE: OK · VERIFY: OK · TESTNET_VERIFY: OK · ADDONS: 1/0/0 · PASS=27 FAIL=0 SKIP=1
Phase 9 — 3 parallel reviewers CRITICAL=0 from all three (MEDIUM 3/2/2, MINOR 1/2/2)
Phase 10 — fix pass No edit applied — 0 critical, 0 in-scope; all 4 medium gaps pre-existing on main, auto-decided with the conservative default and recorded. All 4 Phase-10a static gates PASS.
Phase 10b — smoke re-probe SMOKE: OK — no regressions, both networks
Phase 11 — final verdict CHANGES REQUESTED (CRITICAL=0 MEDIUM=4 MINOR=1) — reported honestly, not upgraded. All 4 MEDIUM are pre-existing on main; the PR's own change drew zero findings.
Working tree hyperliquid.json unmodified by this run — there is no fix-pass edit for the pipeline's commit step to pick up.
Disabled APIs 0 distinct methods, 0 collections. Ledger: (none — no disabled APIs).

Auto-decisions taken at STOP-and-ask gates (unattended run)

  1. Phase 10 fix gate → conservative default: apply nothing. All 4 gaps were mechanically verified byte-identical to main, so none was introduced here; gap Disable data reliability feature #1 would exclude every upstream at boot if the stock pruning template were copied, and gaps feat: Strknt spec v9 #2–Review specs #3 lack the docs needed to apply them correctly. Pushing them would have silently rewritten a production mainnet and testnet spec under a PR scoped to one write method. Carried to the watch-list in the Phase 10 comment instead.
  2. Archive/pruning/GET_EARLIEST_BLOCK triplet gate (normally a Phase 6 STOP; Phase 6 is outside this span) → conservative default: report, do not fix. The triplet is mixed (archive ✓, GET_EARLIEST_BLOCK ✓, pruning ✗) and pre-existing. Removing the archive extension would break archive routing that Phase 8 and 10b both proved working; adding an uncalibrated pruning verification would fail provider startup. Both actions are riskier than reporting.
  3. ws transport → no STOP needed. The spec declares zero FUNCTION_TAG_SUBSCRIBE directives across the resolved import closure, so no ws-capable node was required. Both URLs were still handshake-tested over wss:// and returned 405 Method Not Allowed; recorded, and correctly not a defect.

Run stats — this pipeline run only (Phases 8–11)

Elapsed:         00h 28m 29s  (first→last logged activity)
Transcripts:     1 main + 7 subagent
API round-trips: 380  (main=49, subagents=331)
Fresh tokens:    input=760  output=210409  (sum=211169)
Cached tokens:   cache_read=24306773  cache_write=1355156
Total billed:    25873098 tokens
Split:           main=4814394  subagents=21058704
Models:          claude-sonnet-5  331 calls  21058704 billed
                 claude-opus-5     50 calls   4950937 billed

These figures cover Phases 8–11 only. This PR's spec edit was authored outside the create-spec pipeline (a focused MAG-4426 fix commit), so there is no Phases 1–7 create_spec run to add to them — unlike a fresh-chain PR, these numbers are the whole pipeline cost for this PR, not one half of it.

To re-run any part: /rerun-probe (Phase 8 only) · /rerun-review (9) · /rerun-fix (10 + 10b) · /rerun-final (11) · /rerun-from <8|9|10|11> to run that phase through the end.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants