fix(data): replace non-ASCII lookalike characters in reference tables

[CodeRabbit-byte]

Replaces 153 occurrences of two non-ASCII characters that have no legitimate use in this data, across 79 lines in 7 reference tables.

character	occurrences	replaced with
U+2011 NON-BREAKING HYPHEN	~105	- (ASCII hyphen)
U+00A0 NO-BREAK SPACE	~63	(ASCII space)

file	lines	characters
references/offers/bridges.csv	37	83
references/offers/services.csv	24	39
references/offers/security.csv	8	14
references/offers/oracles.csv	5	7
references/offers/platforms.csv	3	7
references/offers/analytics.csv	1	1
references/providers/providers.csv	1	2

This is not a cosmetic change. The clearest case: 28 assetTypes cells in bridges.csv spell the token standard with a non-breaking hyphen — ERC‑20 rather than ERC-20. Every other row in the database uses the ASCII form. A consumer filtering assetTypes for "ERC-20" matches the ASCII rows and silently skips these 28, with no error and nothing in a visual diff to show why: the two characters are visually identical in almost every font.

The same applies to values like Cross‑chain Messaging in technology, Intent‑based in additionalFeatures, and On‑chain Usage Metrics in dataSources. Each is one grep away from being invisible.

U+00A0 is the same class of problem in whitespace: values such as ["NFT Sales Data␣API"] and 0.5 ETH␣ carry a no-break space where a normal space or nothing was meant. It survives .strip() in several languages' default behaviour, so a trailing one quietly defeats trimming and numeric parsing.

Deliberately left alone. Em dashes (U+2014, 12 occurrences) and en dashes (U+2013, 3) are untouched. They occur in prose description fields where they are plausibly intentional typography, and converting them would be an editorial change rather than a correction. Happy to normalise them too if you would rather the data were pure ASCII throughout — that is a house-style decision, not a defect.

Scope and safety. references/ only, 79 lines — the 11 affected lines under listings/specific-networks/ follow in a separate PR so both stay inside the 100-line limit. A scripted diff against upstream/main confirms every changed cell differs from its original solely by these two substitutions: no column width changes, no line-count change, and no other edit. Where removing a no-break space left a doubled or trailing space inside a value, that value is also trimmed. Zero occurrences of either character remain in these files.

validate_csv.py from the json-tools branch: All checks passed.

If it is useful, I can also contribute a validator rule rejecting U+2011 and U+00A0 in future submissions, so this does not accumulate again — it is the kind of thing that is invisible in review and easy to paste in from a vendor page.

Reward address: 0x1589423BeCC3F87EA9406155EaF4D5E04C34Dcf1 (USDC/USDT, Ethereum mainnet)

[github-actions]

Summary

Status	Count
🔍 Total	111
🔗 Unique	111
✅ Successful	109
⏳ Timeouts	1
🔀 Redirected	13
👻 Excluded	0
❓ Unknown	0
🚫 Errors	1
⛔ Unsupported	0

Errors per input

Errors in ./references/offers/bridges.csv

• [ERROR] https://routernitro.com/swap (at 52:1) | Connection failed. Check network connectivity and firewall settings

Timeouts per input

Timeouts in ./references/offers/bridges.csv

• [TIMEOUT] https://celerscan.com/ (at 13:1) | Request timed out

Full Github Actions output

[USS-Supervisor]

Verdict: REQUEST_CHANGES
Risk: MEDIUM
Summary: The Unicode-normalization part of this PR looks mechanically correct: I parsed all seven changed CSVs, row counts/headers/widths stayed stable, git diff --check is clean, and the touched cells differ only by the stated U+2011 / U+00A0 substitutions plus whitespace cleanup. However, the current review-cycle link-check report found a blocking link issue in a row this PR modifies.

Findings:

• MEDIUM references/offers/bridges.csv, routerprotocol row, actionButtons: the current-cycle link-check run reported https://routernitro.com/swap as an error, and an independent check also fails DNS resolution for routernitro.com. Because this PR touches the routerprotocol reference row, please replace that Bridge URL with a working official Router Protocol/Nitro URL or remove it if no working bridge URL is currently available. I also saw the run time out once on https://celerscan.com/, but that URL returned HTTP 200 in a follow-up check, so I am not treating it as blocking.

Confidence: HIGH

[CodeRabbit-byte]

Fixed. The routerprotocol Bridge link now points at https://app.routerprotocol.com/.

Confirming your finding independently: routernitro.com does not resolve at all — not a 404 but a DNS failure, and the same for www.routernitro.com and app.routernitro.com. The domain is gone rather than the path having moved.

For the replacement I used the URL behind every Launch App button on routerprotocol.com itself rather than guessing at a path. Worth noting that https://app.routerprotocol.com/swap and /bridge both return 404, so the bare app URL is the correct target — the app resolves the bridge and swap views client-side. The row's Docs link at docs.routerprotocol.com still returns 200 and is untouched.

I also grepped the whole repository: no other row references routernitro, so this was the only occurrence.

On celerscan.com — agreed, and thank you for re-checking rather than blocking on it. It returned HTTP 200 for me too on a follow-up probe, so it looks like a transient timeout in the run rather than a dead link.

One note on scope, since this PR was otherwise a pure character substitution: this is the only link edit in it, made because you asked and because the row was one this PR already touches. Everything else remains the U+2011 / U+00A0 normalisation described in the description.

Reward address: 0x1589423BeCC3F87EA9406155EaF4D5E04C34Dcf1 (USDC/USDT, Ethereum mainnet)

[CodeRabbit-byte]

The requested change was pushed on 2026-08-29 and this PR has been waiting on a re-review since.

To restate what was asked and what was done: the finding was that references/offers/bridges.csv, routerprotocol row, linked https://routernitro.com/swap, which the link-check reported as an error and which also failed DNS resolution independently.

That Bridge URL is now https://app.routerprotocol.com/, Router Protocol's own application host. Re-checking both just now:

• https://routernitro.com/ — no DNS resolution, curl exit 6. The original finding still holds.
• https://app.routerprotocol.com/ — HTTP 200.
• https://docs.routerprotocol.com/ — HTTP 200, unchanged and untouched.

The branch was also rebuilt on current main at that time, because it had gone genuinely conflicting on bridges.csv and security.csv while GitHub still displayed it as mergeable. It is now a single commit and shows as mergeable, with the required validation passing.

Nothing else changed: the rest of the PR is still only the stated U+2011 to - and U+00A0 to space substitutions, with row counts, headers and column widths stable.

Ready for re-review whenever the cycle allows.

[github-actions]

Summary

Status	Count
🔍 Total	111
🔗 Unique	111
✅ Successful	111
⏳ Timeouts	0
🔀 Redirected	14
👻 Excluded	0
❓ Unknown	0
🚫 Errors	0
⛔ Unsupported	0

Full Github Actions output

[USS-Supervisor]

Verdict: APPROVE
Risk: LOW
Summary: The updated reference-table normalization now clears the previous broken Router Protocol link finding. All seven touched reference CSVs parse with stable row widths and no duplicate slugs, required validation is passing, and the current-cycle link-check run 33490784047 completed successfully with 111/111 links and 0 errors, timeouts, or unsupported URLs.
Findings: None.
Confidence: HIGH