agentleFS
hyperlane-xyzhyperlane-xyz/hyperlane-monorepo/.agents/skills/investigate-merkle-root-mismatch/SKILL.md

investigate-merkle-root-mismatch

Skill76 starsChanged 20 days agoApache-2.0View on GitHub

The relayer maintains a local merkle tree built from message IDs. When a validator signs a checkpoint, the relayer needs to generate a merkle proof. If the relayer's tree has incorrect message IDs, the roots will mismatch and message delivery will fail. Most likely cause: The relayer indexed incorrect message IDs (possibly due to RPC issues or reorgs), while validators have the correct data. Query Grafana to confirm the mismatch metric is firing: If value is 1, the mismatch is…

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 8 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈8

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If value is 1, the mismatch is confirmed.

    line 57
  2. Very specific:

    If domain_id was not provided, fetch from the registry:

    line 75
  3. Distinctive:

    Check if port 9090 is already in use:

    line 83
  4. Distinctive:

    If not in use, start port-forward in background:

    line 89
  5. Distinctive:

    Wait a few seconds for the port-forward to establish, then verify it's working:

    line 95
  6. Distinctive:

    Start at the latest index - if mismatch, go to 50% of that index

    line 131
  7. Distinctive:

    If match at 50%, mismatch is between 50%-100% - try 75%

    line 132
  8. Distinctive:

    If mismatch at 50%, mismatch is between 0%-50% - try 25%

    line 133
  9. Distinctive:

    Note about fixing: ALL entries from the first mismatch to the latest must be fixed because the merkle tree is cumulative - each root …

    line 206
  10. Distinctive:

    Binary search takes too long: Use larger jumps initially (e.g., 10000 indices)

    line 244
  11. Very specific:

    Shell script errors: Use manual curl commands instead of the bash scripts in rust/scripts/

    line 245
  12. Very specific:

    After investigation, use /fix-merkle-root-mismatch to apply the fixes.

    line 249

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Investigate Merkle Root Mismatch

When to Use

  1. Alert-based triggers:
    • Alert mentions "merkle root mismatch"
    • GCP logs show: "checkpoint root does not match canonical root from merkle proof"
    • The hyperlane_merkle_root_mismatch metric is firing
  2. User request triggers:
    • "Debug merkle tree issues for [chain]"
    • "Investigate the merkle root mismatch on [chain]"
    • "Why is the relayer's merkle tree wrong for [chain]?"

Input Parameters

ParameterRequiredDefaultDescription
originYes-The origin chain with merkle root mismatch (e.g., paradex, ethereum)
domain_idNo-Domain ID of the chain (auto-derived from registry if not provided)
environmentNomainnet3mainnet3 or testnet4

Problem Overview

The relayer maintains a local merkle tree built from message IDs. When a validator signs a checkpoint, the relayer needs to generate a merkle proof. If the relayer's tree has incorrect message IDs, the roots will mismatch and message delivery will fail.

Most likely cause: The relayer indexed incorrect message IDs (possibly due to RPC issues or reorgs), while validators have the correct data.

Prerequisites

  • kubectl access to the relayer pods
  • Grafana MCP server configured

Investigation Workflow

Step 1: Confirm the Alert

Query Grafana to confirm the mismatch metric is firing:

Use mcp__grafana__query_prometheus with:
- datasourceUid: grafanacloud-prom
- expr: hyperlane_merkle_root_mismatch{origin="[origin]"}
- startTime: now-1h
- queryType: instant

If value is 1, the mismatch is confirmed.

Step 2: Get Latest Tree Insertion Index

Query Grafana for the current tree size:

Use mcp__grafana__query_prometheus with:
- datasourceUid: grafanacloud-prom
- expr: hyperlane_latest_tree_insertion_index{origin="[origin]", hyperlane_deployment="[environment]"}
- startTime: now-1h
- queryType: instant

This gives you the latest leaf index to work backwards from.

Step 3: Get Domain ID

If domain_id was not provided, fetch from the registry:

curl -s "https://raw.githubusercontent.com/hyperlane-xyz/hyperlane-registry/main/chains/[origin]/metadata.yaml" | grep domainId

Step 4: Establish Port-Forward to Relayer

Check if port 9090 is already in use:

lsof -i :9090

If not in use, start port-forward in background:

kubectl port-forward omniscient-relayer-hyperlane-agent-relayer-0 9090:9090 -n [environment] &

Wait a few seconds for the port-forward to establish, then verify it's working:

curl -s "localhost:9090/merkle_tree_insertions?domain_id=[domain_id]&leaf_index_start=0&leaf_index_end=1"

Step 5: Binary Search for First Mismatch

Compare validator checkpoints (from S3) against relayer merkle proofs. Use binary search to find the FIRST mismatched index.

Validator checkpoint URL pattern:

https://hyperlane-[environment]-[origin]-validator-0.s3.us-east-1.amazonaws.com/checkpoint_[index]_with_id.json

Comparison function:

# Check a specific index
index=[INDEX]
domain_id=[DOMAIN_ID]
origin=[ORIGIN]
environment=[ENVIRONMENT]

validator_root=$(curl -s "https://hyperlane-${environment}-${origin}-validator-0.s3.us-east-1.amazonaws.com/checkpoint_${index}_with_id.json" | jq -r '.value.checkpoint.root')
relayer_root=$(curl -s "localhost:9090/merkle_proofs?domain_id=${domain_id}&leaf_index=${index}&root_index=${index}" | jq -r '.root')

echo "Index $index:"
echo "  Validator: $validator_root"
echo "  Relayer:   0x$relayer_root"
if [ "$validator_root" = "0x$relayer_root" ]; then echo "  ✓ Match"; else echo "  ❌ MISMATCH"; fi

Binary search strategy:

  1. Start at the latest index - if mismatch, go to 50% of that index
  2. If match at 50%, mismatch is between 50%-100% - try 75%
  3. If mismatch at 50%, mismatch is between 0%-50% - try 25%
  4. Continue narrowing until you find the exact first mismatch index (where index N-1 matches but index N mismatches)

Step 6: Identify Mismatched Message IDs

Once you found the first mismatch index, compare message IDs:

Get validator message IDs:

for i in $(seq [first_mismatch] [first_mismatch + 10]); do
  msg=$(curl -s "https://hyperlane-[environment]-[origin]-validator-0.s3.us-east-1.amazonaws.com/checkpoint_${i}_with_id.json" | jq -r '.value.message_id')
  echo "$i: $msg"
done

Get relayer message IDs:

curl -s "localhost:9090/merkle_tree_insertions?domain_id=[domain_id]&leaf_index_start=[first_mismatch]&leaf_index_end=[first_mismatch + 10]" | jq -r '.merkle_tree_insertions[] | "\(.leaf_index): \(.message_id)"'

Step 7: Get Block Timestamp for Context

Get the block number and timestamp of the first mismatch to understand when the issue started:

# Get block number from relayer
curl -s "localhost:9090/merkle_tree_insertions?domain_id=[domain_id]&leaf_index_start=[first_mismatch]&leaf_index_end=[first_mismatch]" | jq '.merkle_tree_insertions[0].insertion_block_number'

For EVM chains, get timestamp:

cast block [block_number] --rpc-url [rpc_url] -j | jq '.timestamp'

For Starknet chains:

curl -s --request POST --url '[rpc_url]' --header 'Content-Type: application/json' --data '{"jsonrpc":"2.0","method":"starknet_getBlockWithTxHashes","params":[{"block_number":[block_number]}],"id":1}' | jq '.result.timestamp'

Convert Unix timestamp to human-readable:

date -r [timestamp] -u '+%Y-%m-%d %H:%M:%S UTC'

Step 8: Report Findings

Present the investigation results with:

  1. Summary table:
ParameterValue
Chain[origin]
Domain ID[domain_id]
Environment[environment]
First Mismatch Index[index]
Latest Index[latest]
Total Entries to Fix[latest - first_mismatch + 1]
Mismatch Started At[block_number] ([timestamp])
  1. Sample of mismatched entries:
Leaf IndexRelayer Message IDValidator Message IDBlock Number
[idx]0x...0x...[block]
  1. Inform the user that to fix this issue, they should run /fix-merkle-root-mismatch.
  2. Note about fixing: ALL entries from the first mismatch to the latest must be fixed because the merkle tree is cumulative - each root depends on all previous leaves.

API Reference

Relayer Endpoints

EndpointMethodParametersDescription
/merkle_tree_insertionsGETdomain_id, leaf_index_start, leaf_index_endList merkle tree insertions
/merkle_proofsGETdomain_id, leaf_index, root_indexGet merkle proof for a leaf

Validator S3 Checkpoint

https://hyperlane-[environment]-[chain]-validator-0.s3.us-east-1.amazonaws.com/checkpoint_[index]_with_id.json

Response structure:

{
  "value": {
    "checkpoint": {
      "merkle_tree_hook_address": "0x...",
      "mailbox_domain": 514051890,
      "root": "0x...",
      "index": 37352
    },
    "message_id": "0x..."
  },
  "signature": { ... }
}

Common Issues

  1. Port-forward disconnects: Re-run the kubectl port-forward command
  2. Validator S3 returns 404: Checkpoint may not exist yet at that index
  3. Binary search takes too long: Use larger jumps initially (e.g., 10000 indices)
  4. Shell script errors: Use manual curl commands instead of the bash scripts in rust/scripts/

Next Steps

After investigation, use /fix-merkle-root-mismatch to apply the fixes.

Runbook Reference

Full runbook: https://www.notion.so/hyperlanexyz/Merkle-Root-Mismatch-26a6d35200d680a2857dcd0b228d4ab7

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.