fix: rebuild token whitelists from studio catalog with complete metadata

[chrisli30]

• fix: rebuild token whitelists from studio catalog

The whitelist loader (core/taskengine/token_metadata.go) keys cache entries on
TokenMetadata.Id (json:"id") but the legacy files used "address", so every
entry parsed with Id="" and overwrote a single cache slot — only 1 of N
tokens actually became reachable. Rebuilt the four chain catalogs from
studio's app/lib/erc20 (the authoritative UI source), keeping just
{id, name, symbol, decimals} and dropping incomplete or duplicate placeholders.

• ethereum.json: 99 -> 30 entries, all complete
• base.json: 134 -> 23 entries, all complete
• base-sepolia.json: 19 -> 4 entries, all complete (incl. real Circle USDC w/ decimals: 6)
• sepolia.json: 6 -> 6 entries, unchanged

Token enrichment now returns proper name/symbol/decimals for whitelisted
tokens — Base Sepolia balance lookups will no longer surface USDC as
"Unknown Token / UNKNOWN".

Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com

• fix: relabel notification errors heading on success runs

When a workflow ends with status=success but a branch condition skipped a
node, the channel formatters were unconditionally rendering the skip
description under "What Went Wrong" — misleading because nothing actually
went wrong. The deterministic plain-text body composer already gates on
status=="failed"; the channel formatters now match that semantic by
switching the heading to "Skipped Steps" when status=="success", and
keeping "What Went Wrong" for failed/error runs.

Applies to all four formatters (telegram, discord, plaintext, email).
Test updated to assert "Skipped Steps" / no "What Went Wrong" for the
success-with-branch-skipped case.

Co-Authored-By: Claude Opus 4.7 (1M context) noreply@anthropic.com

[claude]

Code Review

Overview

Two independent bug fixes:

1. Token whitelist cache bug — JSON files used \"address\" as the field name, but TokenMetadata has json:\"id\". Every token parsed with Id=\"\", so all N entries collided on the same cache key and only the last one survived. The fix renames the field to \"id\" across all four chain catalogs and takes the opportunity to remove placeholder/malformed entries.

2. Notification heading mislead — When a workflow succeeded but a branch condition skipped a node, all four channel formatters (Telegram, Discord, Plaintext, Email) rendered the skipped-node description under "What Went Wrong", which implies failure. The fix switches to "Skipped Steps" when status == \"success\".

Both fixes are correct and clearly motivated.

---

Token Whitelist Fix

Address case normalization is safe.
token_metadata.go:194 normalises on write (strings.ToLower(token.Id)) and GetTokenMetadata:217 normalises on lookup — so the mixed-case EIP-55 checksummed addresses in the new JSON files (e.g. 0x7Fc66500c84A76Ad7e9c93437bFc5Ac33E2DDaE9) will resolve correctly. No latent case-sensitivity bug here.

Malformed placeholder addresses correctly removed.
The old base.json contained obviously synthetic addresses like 0xd4d042C6c716c0d6C2a5B7F5b7c8C8E6e6b8b8b8 (Maple Finance) and 0x5a6b6a9c7826e2b4b8b8b8b8b8b8b8b8b8b8b8b8 (MediBloc). Removing these is correct — they would never resolve on-chain.

Minor inconsistency: WETH name in ethereum.json.
The new WETH entry uses \"name\": \"WETH\" (the symbol) while every other entry and every other chain file uses a descriptive name (\"Wrapped Ether\"). The value is technically valid but inconsistent with the rest of the catalog.

// ethereum.json — should probably be:
{
  "name": "Wrapped Ether",   // not "WETH"
  "symbol": "WETH",
  "decimals": 18,
  "id": "0xC02aaA39b223FE8D0A0e5C4F27eAD9083C756Cc2"
}

sepolia.json is a no-op except field rename + reorder.
The six tokens and their addresses are unchanged — only \"address\" → \"id\" and a trailing-newline fix. Confirmed correct.

---

Notification Heading Fix

Logic is correct and applied consistently.
All four formatters gate on s.Status == \"success\", which is the right predicate: branch skips only appear during successful runs. Failed/error runs keep "What Went Wrong".

No XSS risk in the email formatter.

sb.WriteString(fmt.Sprintf(`<h3 style="...">%s</h3>`, heading))

heading is always one of two string literals, never user-supplied data — safe.

Test gap: Discord, Plaintext, and Email formatters are not tested for this behaviour.
The updated test in TestFormatTelegramFromStructured_PRDFormat (lines 859–880) asserts \"Skipped Steps:\" and not \"What Went Wrong:\" for the success-with-errors case, which is good. But summarizer_format_test.go has no equivalent tests for Discord, Plaintext, or Email. The logic is identical in all four so the risk is low, but a copy-paste error or future regression in one of those formatters would go undetected. Consider adding parallel table-driven cases (even minimal ones) for the other three.

Minor style note.
The fmt import was added to Discord, Plaintext, and Telegram solely for fmt.Sprintf on a two-part string. String concatenation avoids the import:

// instead of:
sb.WriteString(fmt.Sprintf("\n**%s:**\n", heading))

// could be:
sb.WriteString("\n**" + heading + ":**\n")

fmt.Sprintf is not wrong here, just slightly heavier than necessary. Not a blocker.

---

Summary

Area	Status
Root cause of cache bug	✅ Correctly identified and fixed
Address normalisation safety	✅ Confirmed safe — loader normalises to lowercase
Malformed addresses removed	✅ Good cleanup
Heading logic (all 4 formatters)	✅ Consistent and correct
Telegram test updated	✅
Discord / Plaintext / Email test coverage	⚠️ Missing for the new heading behaviour
WETH name inconsistency in ethereum.json	🔹 Minor (cosmetic only)

The two bugs are real and the fixes are sound. The main follow-up worth doing is adding at least one test case per untested formatter for the success + Errors path.

[Copilot]

Pull request overview

This PR fixes token whitelist loading by rebuilding chain token catalogs to use the id field expected by TokenMetadata.Id, and updates notification channel formatters to avoid “What Went Wrong” framing on successful runs where steps were only skipped by branch conditions.

Changes:

• Rebuilt token whitelist JSON catalogs (Ethereum/Base/Sepolia/Base Sepolia) to use {id, name, symbol, decimals} so whitelist cache keys are populated correctly.
• Updated Telegram/Discord/plaintext/email structured-summary formatters to render the skipped-steps section under “Skipped Steps” when status=="success".
• Updated Telegram formatter test expectations for the success-with-branch-skipped scenario.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
token_whitelist/sepolia.json	Migrates entries from address to id and refreshes token set ordering/content.
token_whitelist/ethereum.json	Rebuilds mainnet token catalog to the {id,name,symbol,decimals} schema used by the loader.
token_whitelist/base.json	Rebuilds Base token catalog to the {id,name,symbol,decimals} schema used by the loader.
token_whitelist/base-sepolia.json	Rebuilds Base Sepolia token catalog to the {id,name,symbol,decimals} schema used by the loader.
core/taskengine/summarizer_format_test.go	Adjusts Telegram formatter test to expect “Skipped Steps” and to forbid “What Went Wrong” on success-with-skips.
core/taskengine/summarizer_format_telegram.go	Switches the errors-section heading based on Summary.Status.
core/taskengine/summarizer_format_plaintext.go	Switches the errors-section heading based on Summary.Status (adds fmt import).
core/taskengine/summarizer_format_email.go	Switches the errors-section heading in analysisHtml based on Summary.Status.
core/taskengine/summarizer_format_discord.go	Switches the errors-section heading based on Summary.Status (adds fmt import).

Comments suppressed due to low confidence (1)

core/taskengine/summarizer_format_email.go:311

• In the email HTML renderer, when Status=="success" the section heading becomes "Skipped Steps", but each entry is still prefixed with a "✗" marker. That marker reads like a failure indicator and undermines the intent of the relabel. Consider switching the per-line glyph based on status (e.g., use a neutral bullet/skip icon when status is success) so success-with-skips doesn’t look like an error list.

		heading := "What Went Wrong"
		if s.Status == "success" {
			heading = "Skipped Steps"
		}
		sb.WriteString(`<div style="margin-bottom: 20px;">`)
		sb.WriteString(fmt.Sprintf(`<h3 style="margin: 0 0 8px 0; font-size: 16px;">%s</h3>`, heading))
		for _, err := range s.Errors {
			sb.WriteString("<p style=\"margin: 0 0 4px 0;\">✗ ")
			sb.WriteString(formatBackticksForChannel(err, "email"))
			sb.WriteString("</p>")

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.