AgentSkillsCN

eta

当用户询问某项任务需要多长时间、请求时间估算，或在开始任何非 trivial 的编码任务之前，立即运行scripts/estimate_task.py，分析代码库的规模范围，并基于数据驱动提供时间估算。在不提问的前提下，清晰展示估算分解、风险因素以及关键的阶段性建议。

SKILL.md
--- frontmatter
name: eta
description: |
  When a user asks how long a task will take, requests a time estimate, or before
  starting any non-trivial coding task, immediately run scripts/estimate_task.py
  to analyze the codebase scope and provide a data-driven time estimate. Show the
  estimate breakdown, risk factors, and checkpoint recommendations without asking.
allowed-tools: |
  bash: python, ls, find, wc
  file: read

ETA

<purpose> Provides realistic time estimates for coding tasks before work begins. Users often underestimate task complexity. This skill analyzes codebase scope, categorizes task complexity, identifies risk factors, and calculates estimates based on measured Claude Code performance baselines. Prevents surprise delays and enables better planning. </purpose>

Instructions

Step 1: Gather Task Context

When triggered, collect:

  • •Task description from the user
  • •Path to relevant codebase (default: current directory)
  • •Number of files likely in scope (if known)

Step 2: Run Estimation

bash
python scripts/estimate_task.py --task "<task_description>" --path "<codebase_path>"

Optional flags:

  • •--files <n>: Override estimated files in scope
  • •--json: Output structured JSON for programmatic use

Step 3: Present Estimate

Return the formatted estimate showing:

  • •Task category (trivial/simple/medium/complex/major)
  • •Scope analysis (files, lines, test coverage)
  • •Time range (low-high estimate)
  • •Breakdown by phase (analysis, implementation, testing, verification)
  • •Risk factors with explanations
  • •Checkpoint recommendations for long tasks

Step 4: Adjust Based on Feedback

If user provides more context:

  • •Re-run with --files override if scope is clearer
  • •Adjust category interpretation based on domain knowledge

NEVER

  • •Start a complex task without providing an estimate first
  • •Give single-point estimates (always provide ranges)
  • •Ignore risk factors when they're detected
  • •Skip the estimate for tasks over 15 minutes
  • •Promise exact completion times

ALWAYS

  • •Run the estimate script before non-trivial tasks
  • •Show the breakdown so users understand where time goes
  • •Flag risk factors visibly with explanations
  • •Recommend checkpoints for tasks over 30 minutes
  • •Update estimates if scope changes mid-task

Examples

Example 1: User asks about task duration

Input: "How long will it take to add user authentication?"

Workflow:

  1. •Run scripts/estimate_task.py --task "Add user authentication" --path .
  2. •Output shows:
    • •Category: Complex
    • •Estimated time: 24-53 minutes
    • •Risk factors: High-risk changes, External dependencies
    • •Recommendation: Break into phases with commits between each

Example 2: Before starting a feature

Input: "Add a dark mode toggle to settings"

Workflow:

  1. •Detect this is a non-trivial task (feature keyword)
  2. •Run scripts/estimate_task.py --task "Add dark mode toggle to settings" --path ./src
  3. •Present estimate: 10-22 minutes (Medium complexity)
  4. •Begin implementation with checkpoint plan

Example 3: Quick fix request

Input: "Fix the typo in the README"

Workflow:

  1. •Detect trivial task (typo keyword)
  2. •Run quick estimate: 3-6 minutes
  3. •Proceed immediately (no detailed breakdown needed for trivial tasks)