Agent Prompt Pack
Use these prompts during ADLC labs. Learners should adapt them, not paste blindly.
Repo Explorer
Section titled “Repo Explorer”Inspect this repository. Do not edit files. Return:- purpose- tech stack- start commands- test commands- main entry points- important data flow- likely risk areas- unknowns that need manual verificationFeature Mapper
Section titled “Feature Mapper”Trace the feature path for POST /tickets. Do not edit files. Show:- HTTP route- request parsing- validation- business logic- database writes- audit behavior- tests- failure modesUse file paths.Implementation Worker
Section titled “Implementation Worker”Implement a bounded change:
Goal:Add ticket priority with allowed values low, normal, high.
Constraints:- Touch only the starter project.- Preserve existing API behavior.- Validate priority server-side.- Default to normal.- Add tests.- Update API contract.
Verification:Run python3 -m unittest -v and report results.Debugging Planner
Section titled “Debugging Planner”The bug is: non-agent users can approve drafts when they should receive 403.Do not suggest a fix yet. Create a debugging plan with reproduction steps,boundary checks, expected logs, and the test that should fail before the fix.Code Reviewer
Section titled “Code Reviewer”Review this diff for bugs, missing tests, security issues, data correctness,API contract mismatches, and operational risks. Findings first. Reference filesand lines. Do not summarize unless there are findings or residual risks.Runbook Reviewer
Section titled “Runbook Reviewer”Review this runbook. Find missing start/stop steps, config assumptions, healthchecks, logs, smoke tests, rollback steps, and common failure recovery paths.Return concrete gaps only.Capstone Critic
Section titled “Capstone Critic”Review this capstone as if it must run for a real user tomorrow. Find gaps inworkflow fit, architecture, data, API, tests, operations, privacy, audit,rollback, and agentic verification. Rank by severity.Prompt Review Checklist
Section titled “Prompt Review Checklist”Before sending a prompt:
- Is the goal specific?
- Is context included?
- Are editable files bounded?
- Is out-of-scope explicit?
- Are verification commands included?
- Is expected response format clear?
- Did you ask for evidence?