Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    real eta

    1

    Data-driven timeline estimation for AI agent projects using multi-clock forecasting and reference-class data.

    Free

    Security scanned
    real eta

    real eta

    Example session with this skill installed

    How long will it take to build a basic Stripe subscription integration for my Next.js app? I need it live by Friday.

    • Read your context and instructions
    • Compiled the real eta
    • Generated the document

    Stripe Live Launch Timeline

    Target launch: Wednesday, 28 October (P90: Friday, 30 October). Friday is a realistic target.

    • Agent work: 2–5 hours
    • Your time: 4–6 hours

    Two factors determine the launch date: how quickly Stripe activates your live account and how long end-to-end payment testing takes. Coding is not the main bottleneck.

    Assumptions

    This estimate assumes an existing Next.js app with login, one or two plans, Stripe Checkout, the Stripe Customer Portal, and a webhook that keeps subscription status in your database.

    A custom payment form, trials, usage-based billing, or tax handling would increase the estimate.

    Implementation Plan

    Phase 0 — Monday–Tuesday: Integration

    • Implement the Checkout route.
    • Build the webhook handler, including signature verification, safe handling of repeated events, and database status updates.
    • Restrict access to paid pages.
    • Add a link to the Stripe Customer Portal.
    • Verify that everything works in test mode.

    Phase 1 — Tuesday–Wednesday: End-to-End Testing

    Use the Stripe CLI to test

    • Signup
    • Renewal
    • Failed card payments
    • Cancellation
    • Plan upgrades

    Use Stripe test clocks to simulate billing periods without waiting for real time to pass. Fix any issues discovered during testing.

    Phase 2 — Wednesday–Thursday: Live Deployment

    • Configure live API keys and environment variables.
    • Set up the live webhook endpoint.
    • Deploy the application.
    • Complete one real card payment and refund it to verify the live payment flow.

    Start Today

    1.Submit your Stripe live-mode activation request. Provide your business details and bank account information. Activation can be quick, but it may take days if Stripe requests additional documents.
    2.Confirm country eligibility. Verify that your business country is supported by Stripe. If it is not, the proposed launch plan will need to change.
    3.Finalize plan names and prices. Allow approximately 15 minutes.

    What Could Delay the Launch?

    1. Stripe activation takes longer than expected.

    Fallback: Ship the integration in test mode by Friday and enable live payments as soon as activation clears.

    2. Webhook edge cases emerge.

    Events may arrive out of order or be delivered more than once. Fallback: Retrieve the current subscription state from Stripe when processing relevant webhooks rather than relying solely on the event payload. Make processing idempotent and account for out-of-order events.

    Why These Estimates?

    The agent-work estimate is my judgment for approximately six work units, each lasting one or two hours.

    -Agent execution: The estimate is informed by METR's reported 80% time horizon of roughly 1–1.5 hours for frontier models (METR, May 2026).
    -Review and rework: The review time and 15–25% rework allowance are informed by Chen & Stratton.
    -Existing application work: Adding a feature to an existing app is most comparable to Mitchell Hashimoto's Ghostty build.

    These are planning estimates, not measured predictions. There is no calibration history yet, and I have not inspected your codebase.

    jellyfish.pdf

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

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

    About this skill

    Software delivery dates are notoriously difficult to predict, especially when mixing AI agents with human review and third-party dependencies. This skill replaces guesswork with a data-driven estimation framework that separates agent performance from human bottlenecks. It helps you calculate realistic timelines by breaking work into build, research, human, and wait units.

    What it does

    • Multi-clock estimation separates agent execution, human review cycles, and external wait times.
    • Reference-class forecasting compares your project against measured builds and evidence-based study data.
    • Confidence modeling provides P50 (likely) and P90 (high confidence) dates rather than single-point estimates.
    • Calibration tracking uses local scripts to log actual versus estimated time, refining future predictions.

    How to execute

    1. Scope the MVP to define the smallest shippable unit, the codebase state, and the model being used.
    2. Decompose the work into discrete units of build, research, human, wait, and unknown tasks.
    3. Calculate three clocks to determine agent hours, human review requirements, and calendar-day waits.
    4. Generate the roadmap with a three-phase plan, risk assessments, and immediate next steps.

    Frameworks & tools

    The skill utilizes a calibrate.py script for local performance tracking and relies on internal evidence.md and reference-classes.md files for baseline metrics.

    Why this beats prompting it yourself

    Standard LLMs hallucinate timelines based on human-written tickets, often overestimating agent speed by 4 to 7 times. This skill forces the agent to respect external wait times like App Store reviews and payment approvals that AI cannot accelerate.

    Use cases

    • MVP planning for new product launches needing realistic shipping dates.
    • Roadmap validation to verify if a user-suggested deadline is mathematically feasible.
    • Capacity planning for developers supervising multiple agents simultaneously.
    • Stakeholder reporting to provide professional P50/P90 confidence intervals.

    Known limitations

    Calibration requires access to a local Python environment. Evidence data older than six months may overestimate agent durations. Agent speed gains are significantly lower in large legacy codebases.

    Calibration requires local Python. Evidence older than 6 months may overstate agent build times. Performance gains are lower in large legacy codebases where review and coordination dominate.

    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 7 days ago

    • Free with an Agensi account

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Environment Variables
    Listed7 days ago

    What's inside

    Frequently Asked Questions

    Standard LLMs often hallucinate timelines based on optimistic ticket text; this skill uses a "multi-clock" framework to account for non-AI factors like human review cycles and external API dependencies that AI cannot speed up.

    The package includes the core estimation logic, the `calibrate.py` script for local tracking, and structured baseline metrics in `evidence.md` and `reference-classes.md`.

    Yes, while the core logic is universal, the included `calibrate.py` script requires a local Python environment to track your specific agent's performance and refine future predictions.

    This skill is agent-agnostic and designed to work with any LLM or coding assistant, as it focuses on the project management framework and data-driven forecasting rather than the specific model execution.

    Instead of a single date, you receive P50 (likely) and P90 (high confidence) estimates, allowing you to present stakeholders with a mathematically grounded range that accounts for potential risks.