Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Antigravity Fast-IO Workflow Checklist

    1

    Use when accelerating scientific coding, dataset inspection, or multi-file refactors in Claude Code, Cursor, Copilot, or Antigravity without context bloat.

    Free

    Security scanned
    Antigravity Fast-IO Workflow Checklist

    Antigravity Fast-IO Workflow Checklist

    Example session with this skill installed

    Audit why our evaluation pipeline in src/eval/ is dropping NaN rows, update the metric key across configs/ and paper/tables.tex, and verify nothing broke without scanning .venv or dumping raw CSVs.

    • Read your context and instructions
    • Compiled the antigravity fast-io workflow

    Applying Fast-IO Bounded Parallel Protocol (3 round-trips total vs. 14 unguided turns):

    1. Pass 1 — Scoped Discovery (Excluded .venv, wandb, results_raw):
      Matched nan_policy & val_macro_f1 across 4 files (src/eval/metrics.py:42, src/eval/runner.py:118, configs/ablation.yaml:19, paper/tables.tex:87).

    2. Pass 2 — Parallel Bounded Read & Root-Cause Diagnosis:
      Read all 4 target spans in parallel (<=160 lines each, zero raw .csv ingestion):

      FileLinesFinding
      src/eval/metrics.pyL35–L92dropna() strips valid sparse-sensor rows before macro-F1 aggregation
      configs/ablation.yamlL1–L45Uses deprecated key val_f1_score instead of val_macro_f1
      paper/tables.texL70–L110\cite{chen2025sparse} intact; macro macro-column macro-header aligned
    3. Pass 3 — 3-Step Verified Patch & Batched Smoke Check:

      • Patched metrics.py (nan_to_num mask) + ablation.yaml in parallel;
      • Verified exact 2-file delta via git diff --stat (0 unintended \cite{} edits) + ran pytest tests/test_metrics.py -q (6 passed in 0.41s).
    lines each, zero raw .csv

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

    About this skill

    Stop losing 10+ minutes per session to fragmented file reads, unfiltered root directory scans, and context-window explosions.

    Antigravity Fast-IO Workflow Checklist equips your coding agent (Claude Code, Cursor, GitHub Copilot, Windsurf, Codex, and Google Antigravity) with a high-velocity execution discipline built for scientific computing, ML engineering, and academic repositories:

    • Bounded Parallel File Reads: Batch independent code/config reads in a single round-trip with explicit 150–200 line windows instead of sequential 50-line guesswork.
    • Scoped Search Hygiene: Exclude .venv, wandb, lightning_logs, pycache, and node_modules by default so repository searches finish in milliseconds.
    • Single-Shot Computation Batching: Combine dataset header checks, metric aggregation, figure generation, and build verification into one deterministic turn.
    • 3-Step Multi-File Refactor Loop & Numerical Integrity: Audit match counts before editing, apply minimal diffs, verify with git diff, and enforce strict non-fabrication rules on p-values, confidence intervals, and DOIs.

    Need deterministic Python CLI enforcement, zero-OOM binary tensor inspection, remote GPU pre-shutdown SHA-256 gates, and <50ms workspace safety hooks? Upgrade to the Sci-Fast-IO Pro: Scientific Agent Suite by @cheat-s-azuki.

    Requires a Python environment for binary data probing. Efficiency gains are limited if the host agent lacks support for parallel tool calls or shell execution.

    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 2 days ago

    • Free with an Agensi account

    Requires

    Git

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    This skill drastically reduces latency and token usage by forcing the agent to batch multiple file reads into single turns and preventing the accidental ingestion of massive binary files or junk directories like node_modules.

    The skill is environment-agnostic and performs exceptionally well with Antigravity, Claude Code, Cursor, and Windsurf, as well as any agent utilizing Agensi Remote MCP environments.

    Yes, it is specifically optimized for scientific workflows, providing the agent with logic to safely probe .parquet, .safetensors, and .h5ad formats using Python snippets rather than raw reads.

    Unlike generic prompting, this skill provides a structured execution protocol for line-bounded windows and multi-file replacements that agents cannot reliably maintain through simple instructions alone.

    The Lite version includes the core logic for task triage, two-turn search, tiered thresholding for logs, and the verified batch edit protocol; updates are pushed automatically through the marketplace.

    The skill requires an environment where the agent has access to a Python interpreter or shell execution to perform its data probing and batching operations effectively.

    New in Research & Analysis