A review command that saves a possible blind-spot finding as a draft outside the project worktree. It first checks whether the issue is actually a correctness, performance, or security defect.

Use it to record a qualified review finding, choose a reusable issue category, and write a timestamped draft for later review.

It prevents ordinary defects from being filed as minor review observations and lets you capture a finding without credentials or network access.

How BenchBox-dev/BenchBox configures Claude Code

Reuse

Borrowing it

Nothing to install: this file belongs to BenchBox-dev/BenchBox. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file

curl -O https://raw.githubusercontent.com/BenchBox-dev/BenchBox/develop/.claude/commands/blind-spot.md

Clone the repo

git clone --depth 1 https://github.com/BenchBox-dev/BenchBox

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/commands/benchbox-dev/benchbox/blind-spot.svg)](https://agentmods.dev/commands/benchbox-dev/benchbox/blind-spot)

Your own site

<a href="https://agentmods.dev/commands/benchbox-dev/benchbox/blind-spot"><img src="https://agentmods.dev/badge/commands/benchbox-dev/benchbox/blind-spot.svg" alt="Measured on agentmods" height="20"></a>

Per session 28 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 769 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.00028	$0.00769
Opus 5	$0.00014	$0.00385
Sonnet 5	$0.00006	$0.00154
Haiku 4.5	$0.00003	$0.00077

Measured 2d ago against content hash e72cbdd831d0, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

blind-spot 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.

.claude/commands/blind-spot.md · 59 lines

How it starts

The opening of the file, as written. The whole thing — 59 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Context

Branch: !git branch --show-current
Date/time: !date +%Y-%m-%d-%H%M%S
Unsynced drafts: !ls ~/.benchbox/finding-drafts/*.md 2>/dev/null | wc -l | tr -d ' '

Your task

Follow docs/agent/review-protocol.md exactly. Apply the defect gate [REVIEW-DEFECT-001] before anything else: a finding that materially affects correctness, performance, or security is a defect, not a blind spot. If the gate triggers, refuse to file and offer the user the TODO / inline-fix / escalation menu instead. If the gate passes:

Pick the slug. Derive a 3–6 word kebab-case slug capturing the class, not the instance. Example: "react keys collide on platform name" → react-key-collision-class. If the user gave nothing, ask one short clarifying question and stop.
Compose the path ~/.benchbox/finding-drafts/YYYY-MM-DD-HHMMSS-<slug>.md using the timestamp from the Context block above (run mkdir -p ~/.benchbox/finding-drafts first). Drafts live outside every worktree — capture needs no credentials and no network, and per [REVIEW-CAPTURE-001] the draft file is the sole in-review write for a finding. Filename stem must equal the id field in the frontmatter. If the path collides, append -2, -3, … before .md.
Write the file using the frontmatter schema and body shape in _project/blind-spots/README.md (storage spec). Required fields are id, date, status, finding_kind, review_context. Optional capture fields: observed_sha and evidence (a list of {path, pattern, note}) for provenance; leave the triage fields (urgency, breadth, confidence) unset — they are assigned later at triage, never required at capture.

Validate:

uv run --project _project/scripts -- python _project/scripts/validate_blind_spot.py ~/.benchbox/finding-drafts/<filename>

If validation fails, fix the frontmatter and re-run.

Report with Recorded: ~/.benchbox/finding-drafts/<filename>.md and a quote of the body. Stop there — do not commit, do not push, do not run make pr-open, and do not write the finding into the tracker database. Per [REVIEW-CAPTURE-001] the draft file is local-only; syncing it into the tracker (todo finding sync) is a separate, user-authorized landing step.

Read the full file on GitHub · 59 lines

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.

2d ago First seen · 59 lines · 28 tokens per session scan A e72cbdd831d0

Subscribe to this mod's changes

blind-spot is a command published in the GitHub repository BenchBox-dev/BenchBox (12 stars, last pushed today), licensed MIT. It adds 28 tokens to every session and 769 once invoked, about $0.0001 per session on Opus 5. 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 2026-09-04.

Code review

BenchBox: Command for Claude Code

Borrowing it

Wrote this? Show the measurements

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

Context

Your task

What this file has done since we first saw it

Other commands, from other repositories

sdd-init

review-branch

init

merge-conflict-analysis

review-sdk-app

repo-audit