deep_read_audit

A protocol for auditing tools, prompts, skills, and rules by reading complete files, examining them in parallel, and recording numbered findings with exact locations and quotes.

In plain words
What is it for?
Use it for end-to-end audits that identify shared code to centralize, confusing descriptions, and priorities for cleanup.
Why use it?
It reduces the chance of missing duplicated, conflicting, or unclear instructions during a codebase review.

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/homenshum/nodebenchai/deep_read_audit
Clone the repo
git clone --depth 1 https://github.com/HomenShum/NodeBenchAI

Made for: Cursor.

Per session 0 Nothing until a file matches its globs; then the whole rule loads.
When invoked 514 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 $0.00514
Opus 5 $0.00000 $0.00257
Sonnet 5 $0.00000 $0.00103
Haiku 4.5 $0.00000 $0.00051

Measured 2d ago against content hash 9bcd6de24eb7, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

deep_read_audit 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/deep_read_audit.mdc · 46 lines

What it actually says

Deep-Read Audit Protocol

When collocating across tools, prompts, skills, or rules — NEVER edit from a section-level read. Full end-to-end reads only.

Protocol

  1. Categorize — Group all target files into 4-6 non-overlapping categories by domain
  2. Parallel subagents — Launch one Explore subagent per category, all simultaneously
  3. Full reads required — Each subagent reads EVERY file in its category end-to-end. No partial reads, no grep-only
  4. Numbered findings — Each finding must include:
    • [N] FILE:LINE — exact location
    • "exact quote of relevant code/text" — verbatim from the file
    • ISSUE: — description of duplication, obscurity, or centralization need
  5. Wait for all — Do not synthesize until ALL subagents complete
  6. Synthesize — Produce a single consolidated report with:
    • P0 (extract immediately), P1 (fix this sprint), P2 (next sprint)
    • Estimated impact per tier
    • Proposed module paths for centralized code

What to look for

  • Duplicated helper functions across files
  • Overlapping tool/rule descriptions that confuse discovery
  • Shared patterns that should be centralized (DB setup, ID generation, timestamps, fetch wrappers)
  • Inconsistent naming conventions (params, response shapes, error formats)
  • Tools/rules that do similar things in different files
  • Hardcoded values that should be shared constants
  • Stale/deprecated code still callable
  • Instructions duplicated across CLAUDE.md, .claude/rules/, .cursor/rules/

Anti-patterns

  • Reading only the first 50 lines and inferring the rest
  • Grepping for keywords instead of reading full context
  • Editing before the audit completes
  • Skipping large files (72KB+ agent files are WHERE the duplication hides)
  • analyst_diagnostic — root cause before fix
  • reexamine_process — orchestrator for when/how to re-examine
  • completion_traceability — cite back to original request
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 · 46 lines · 0 tokens per session scan A 9bcd6de24eb7

Subscribe to this mod's changes

deep_read_audit is a cursor rule published in the GitHub repository HomenShum/NodeBenchAI (14 stars, last pushed 18d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 514 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.