blindspot

blindspot is a skill for Claude Code, Codex from vinta/hal-9000. It costs 90 tokens per session (906 once invoked), scanned A, original, MIT.

A questioning guide for people entering an unfamiliar tool, domain, or codebase. It finds missing assumptions and questions the user may not know to ask.

In plain words
What is it for?
Use it before working in an unfamiliar area to map unknowns, identify relevant facts, and prepare focused questions.
Why use it?
It reduces the risk of starting with hidden misunderstandings or overlooking important decisions when the subject is new.

Skill for Claude CodeCodex

Part of the skills plugin — 10 skills shipped together

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 skills/vinta/hal-9000/blindspot
Any agent
npx skills add vinta/hal-9000 --skill blindspot
Clone the repo
git clone --depth 1 https://github.com/vinta/hal-9000

Made for: Claude Code, Codex.

Or install skills, the plugin that ships this one along with the rest of its 10 skills.

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.

agentmods badge for blindspot

README.md
[![agentmods](https://agentmods.dev/badge/skills/vinta/hal-9000/blindspot.svg)](https://agentmods.dev/skills/vinta/hal-9000/blindspot)
Your own site
<a href="https://agentmods.dev/skills/vinta/hal-9000/blindspot"><img src="https://agentmods.dev/badge/skills/vinta/hal-9000/blindspot.svg" alt="Measured on agentmods" height="20"></a>
Per session 90 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 906 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 $0.00090 $0.00906
Opus 5 $0.00045 $0.00453
Sonnet 5 $0.00018 $0.00181
Haiku 4.5 $0.00009 $0.00091

Measured 4d ago against content hash 25d1c1abedef, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

blindspot 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 4d 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.

skills/blindspot/SKILL.md · 58 lines

How it starts

The opening of the file, as written. The whole thing — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Blindspot

The user is about to work in territory they don't know. Two kinds of blindness live there: unknown unknowns (questions they don't know exist) and unknown knowns (assumptions too obvious to write down, and things they're sure of that are wrong). Convert the first into known unknowns, name the second, then hand back a map they can prompt with.

Boundaries: best-practices returns recommendations — this skill returns questions. Grilling stress-tests decisions the user can already defend — this skill maps territory where they can't answer yet. Neither a tutorial nor a plan.

Workflow

1. Surface the framing

Before asking anything, state as bullets: the assumptions the request takes for granted, and the missing information that would change the approach. This names the user's unknown knowns up front. If their goal or familiarity is still unclear, fold one calibration question into the first interview round.

2. Recon

Facts are your job, never the user's. Sweep before asking:

  • Repo: Explore agent for existing patterns, conventions, and adjacent solutions.
  • Tools and domain: find-docs for current APIs and config; WebSearch for pitfalls ("X gotchas", "X common mistakes") — pitfalls live in issue threads and post-mortems, not getting-started docs.

Recon output is question fuel, not findings: each item becomes a question only the user can answer, an assumption to name, or nothing. A user who wants recommendations instead of questions wants the best-practices skill.

3. Interview

AskUserQuestion rounds, up to 4 questions each, 2–3 rounds total.

  • Ask where importance is high and evidence is low: architecture-changers and behavior-definers recon couldn't settle. Skip polish.
  • Anchor in the user's concrete past ("last time you did X, what happened?"), not hypotheticals — people speculate confidently and wrongly.
  • Give each option a trade-off description; put your recommended option first, labeled "(Recommended)". The built-in "Other" is the escape hatch.
  • Include one premortem question phrased in past tense — "it's three months later and this failed: what broke?" — past tense recruits prospective hindsight; "what could go wrong" is measurably weaker.
  • "I don't know" is a first-class answer: record it as a known unknown and move on.

Read the full file on GitHub · 58 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.

  1. 4d ago First seen · 58 lines · 90 tokens per session scan A 25d1c1abedef

Subscribe to this mod's changes

blindspot is a skill published in the GitHub repository vinta/hal-9000 (125 stars, last pushed 7d ago), licensed MIT. It adds 90 tokens to every session and 906 once invoked, about $0.0005 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 2026-08-30.

Related

Other skills, from other repositories

configure-nightwatch

Configures Laravel Nightwatch data collection, sampling rates, filtering rules, and redaction policies. Use when setting up Nightwatch, managing data volume, protecting sensitive data (PII), or optimizing event collection for production workloads.

coollabsio/coolify · 49 tokens

laravel-actions

Build, refactor, and troubleshoot Laravel Actions using lorisleiva/laravel-actions. Use when implementing reusable action classes (object/controller/job/listener/command), converting service classes/controllers/jobs into actions, orchestrating workflows via faked actions, or debugging action entrypoints and wiring.

coollabsio/coolify · 62 tokens

debugging-output-and-previewing-html-using-ray

Use when user says "send to Ray," "show in Ray," "debug in Ray," "log to Ray," "display in Ray," or wants to visualize data, debug output, or show diagrams in the Ray desktop application.

coollabsio/coolify · 58 tokens

fortify-development

ACTIVATE when the user works on authentication in Laravel. This includes login, registration, password reset, email verification, two-factor authentication (2FA/TOTP/QR codes/recovery codes), passkeys, profile updates, password confirmation, or any auth-related routes and controllers. Activate when the user mentions…

coollabsio/coolify · 199 tokens

pest-testing

Use this skill for Pest PHP testing in Laravel projects only. Trigger whenever any test is being written, edited, fixed, or refactored — including fixing tests that broke after a code change, adding assertions, converting PHPUnit to Pest, adding datasets, and TDD workflows. Always activate when the user asks how to…

coollabsio/coolify · 171 tokens

configuring-horizon

Use this skill whenever the user mentions Horizon by name in a Laravel context. Covers the full Horizon lifecycle: installing Horizon (horizon:install, Sail setup), configuring config/horizon.php (supervisor blocks, queue assignments, balancing strategies, minProcesses/maxProcesses), fixing the dashboard…

coollabsio/coolify · 137 tokens