Claude Code Prompts

Use these prompt starters when working in a repo-aware terminal agent surface.

Prompt Objective

Get high-context repository reasoning, safer patch planning, and focused validation steps.

When to Use

  • mapping a repo before edits
  • reviewing diffs
  • planning or applying cross-file fixes

Required Context

  • target files or directories
  • expected validation commands
  • any do-not-touch areas

Prompt Starters

Map this repository for a new engineer. Focus on entry points, build/test commands, and risky directories. Do not edit yet.

Review the current diff like a strict senior engineer. List findings first with severity, file path, and missing tests.

Implement the smallest safe fix for this bug. Add one focused regression test and list rollback steps.

Variables to Customize

  • task objective
  • validation commands
  • risk areas

Example Filled Prompt

Review the current docs diff like a strict senior engineer.
Focus on broken links, anchor regressions, and missing validation updates.
List findings first and cite the affected file for each finding.

Example Expected Output

  • ordered findings
  • file paths
  • one narrow validation step per finding

Weak Prompt Versus Strong Prompt

  • Weak: check this repo
  • Strong: define scope, risks, output format, and validation expectations

Failure Patterns

  • asking for edits before context gathering
  • not naming the validation command

Tool-Specific Adjustments

  • give file paths or directories explicitly when the repo is large
  • ask for a plan before patching on risky changes

Human Review Checklist

  • confirm the agent inspected the right files
  • validate high-risk claims manually

Security Considerations

State if shell execution, writes, or network access should be avoided for the task.