Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Blindspot

    1

    Catches the missing context in a vague request before Claude guesses, with a 5-second confidence check and an escape hatch.

    Free

    Security scanned
    Blindspot

    Blindspot

    Example session with this skill installed

    Plan our team offsite.

    • Read your context and instructions
    • Compiled the blindspot

    Context Confidence: 35%

    1. How many people, and what's the budget? (+25%)
    2. What's the goal: strategy, team bonding, or both? (+20%)
    3. Where and when: a local day or travel? (+12%)

    Answer any, or say go to run now assuming: 12 people; moderate budget; team bonding; local, one day.


    User: 25 people, about $15k, mostly strategy for next year with some bonding. Two days, we can travel within driving distance.

    Claude (confidence now 90%, so it answers in full):

    Two-Day Strategy Offsite (25 people, ~$15k)

    Venue: a retreat lodge within 2 hours' drive, with one large room plus two breakout rooms. Budget about $7,500 for lodging and meeting space.

    Day 1: Look back, look out

    • Morning: last year's wins and misses, then a 90-minute customer and market review
    • Afternoon: three breakout groups each draft one 2027 priority
    • Evening: group dinner and a low-key activity (cooking class or guided hike), about $2,500

    Day 2: Commit

    • Morning: groups present; the whole team ranks priorities with dot voting
    • Midday: owners and first 30-day milestones for the top three priorities
    • Afternoon: a short retro, then head home by 4pm

    Budget: lodging and rooms $7,500 · food $3,500 · activity $2,500 · facilitation supplies and buffer $1,500

    Next steps: book the venue this week, send a pre-read 7 days out, and name breakout leads.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Vague prompt in, generic answer out. When you ask for a go-to-market plan or a fix for flaky deploys with no context, an AI quietly fills the gaps with safe, average guesses. The answer looks complete, and you never see which assumptions were made.

    Blindspot fixes that. On complex, high-stakes asks that are missing real context, Claude pauses before answering for a 5-second Context Check.

    What the check shows

    • Context Confidence: a score for how much of what matters you've provided.
    • Up to 3 questions: only the highest-leverage ones, each with how much it raises the score.
    • The escape hatch: say "go" at any time, and Claude shows the exact assumptions it will run with.

    Those default assumptions are the point. Seeing them is usually the moment you realize what you left out.

    When it fires

    • Strategy, plans, drafts, architecture, analysis and troubleshooting, when confidence is below 85%.
    • When you ask: "check my blindspots," "scope this prompt," "rate my context."

    Designed not to be annoying

    • Never fires on factual questions, quick rewrites, syntax help, casual chat, or prompts that are already clear.
    • At most 3 questions per round and two rounds total, then it runs with its best defaults.
    • At 85% confidence, or the moment you say "go," it delivers the full answer with no extra chatter.
    • No lectures about prompting.

    Pairs with Skill Spotter

    Once Blindspot helps you dial in a task, Skill Spotter can package that setup as a reusable skill, so you never answer the same questions twice.

    Free and MIT-licensed. Source: https://github.com/BrianChase515/blindspot

    Limited to two rounds of questions per task. Does not trigger for tasks already rated above 85% confidence or simple factual lookups.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 10 days ago

    • Free with an Agensi account

    Permissions

    No special permissions declared or detected

    Listed10 days ago

    What's inside

    Frequently Asked Questions

    This skill identifies hidden assumptions and missing variables in complex prompts, preventing generic outputs by requiring a high-leverage context check before the LLM generates a full response.

    It is a logic-layer utility optimized for Claude Code, Cursor, and the Claude web/mobile interface, working as a meta-prompting layer within your existing development or business workflows.

    Unlike generic 'ask me questions' prompts, this skill enforces a strict three-question limit and provides an immediate 'Assumption Map' so you can see exactly how the agent will fill in the blanks if you choose to skip the check.

    The skill includes 'Efficiency Throttling,' which means it automatically bypasses the check for simple factual queries or syntax fixes, only interrupting when it detects a high-stakes, vague request.

    The purchase includes the core logic-layer prompt files and a setup guide to integrate the skill into your preferred AI interface or coding environment.

    To prevent 'loop fatigue,' the skill is hard-coded to stop asking for context after two rounds, ensuring you never get stuck in a recursive questioning cycle.

    New in Prompt & Skill Engineering