text-first-normalizer

A rule for converting provider responses into readable plain text for compliance checks and audit records. Token counts and other detailed usage data are optional for consumer-facing connections.

In plain words
What is it for?
Use it when building normalizers for ChatGPT Web, Claude Web, Cursor, Gemini Web, or similar consumer interfaces, and when handling extraction errors.
Why use it?
It keeps adapters working when chat websites change their response formats and avoids requiring fragile, complete data conversions.

Cursor rule for Cursor

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add rules/alphabitcore/nexus-gateway/text-first-normalizer
Clone the repo
git clone --depth 1 https://github.com/AlphaBitCore/nexus-gateway

Made for: Cursor.

Per session 0 Nothing until a file matches its globs; then the whole rule loads.
When invoked 450 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00450
Opus 5 $0.00000 $0.00225
Sonnet 5 $0.00000 $0.00090
Haiku 4.5 $0.00000 $0.00045

Measured 2d ago against content hash 2a168b810f0c, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

text-first-normalizer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.cursor/rules/text-first-normalizer.mdc · 36 lines

What it actually says

Text-first normalizer (binding)

For consumer-surface traffic (chatgpt-web, claude-web, cursor, gemini-web, and any future consumer-side wire format), the normalizer's only required output is readable text. Losing token / usage stats at this stage is acceptable.

Canonical memory: feedback_compliance_proxy_text_first.

What's required

ExtractText(raw []byte) (Extracted, error) MUST produce the user's prompt + the assistant's response as plain UTF-8 text. Hooks evaluate against this text; audit captures this text. If text extraction fails, the request still flows but is recorded with extract_error=....

What's optional

Token counts, cost stats, role-by-role decomposition, tool-call structure — nice-to-have. Consumer wire formats are inconsistent and brittle; insisting on full canonical structure produces fragile adapters that break on every minor provider UI update.

For API-surface traffic (the same provider hit via SDK / /v1/*), the same adapter typically produces both text AND structured usage; that's a bonus, not a contract.

What this rule prevents

Adapter authors writing 200-line OpenAI-shape canonicalizers for claude-web just to satisfy "completeness". Brittle structural normalizers break on every consumer-surface UI update; keep adapters text-first.

Tier-2 NonJSONDetector

For non-JSON wire formats (binary protocols, multipart, gRPC-Web, raw audio): add a NonJSONDetector in packages/shared/transport/normalize/extract/detector.go. Tier-1 adapters delegate to the detector. Do NOT write a fresh per-host adapter for a new non-JSON format. Canonical memory: feedback_tier2_nonjson_detector_framework.

Skipping this rule requires explicit user approval in chat.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 36 lines · 0 tokens per session scan A 2a168b810f0c

Subscribe to this mod's changes

text-first-normalizer is a cursor rule published in the GitHub repository AlphaBitCore/nexus-gateway (22 stars, last pushed 7d ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 450 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.