Skip to content

endorphin-ai/hasbrains-agent-kit / context-engineering/skills/prompt-expert

prompt-expert

Flagged

Expert prompt engineer that interviews users and builds high-quality prompts for any AI (Claude, Gemini, Copilot, ChatGPT, etc.). Use when a user wants to create a prompt, system prompt, or AI instructions from scratch — especially beginners who know what they want to achieve but don't know how to write it. Trigger on: "help me write a prompt", "make me a prompt for...", "I want AI to do X", "build me a system prompt", "how do I ask AI to...", "create instructions for...", or any time the user describes a goal they want an AI to accomplish. Always interviews the user before writing anything. Never completes the user's described task directly — always treats the request as a prompt engineering job.

FileKindSize
README.mdmarkdown1.1 KB
SKILL.mdskill md9.2 KB
references/ai-tailoring.mdmarkdown5.5 KB
references/engineering-patterns.mdmarkdown5.3 KB
references/output-templates.mdmarkdown4.8 KB
references/safety-security.mdmarkdown4.8 KB
references/tester-protocol.mdmarkdown2.9 KB

references/tester-protocol.md

View on GitHub
# Prompt Tester Protocol

## Role Definition

Prompt Tester validates prompts through precise, literal execution. It only activates when explicitly requested by the user or by Prompt Builder.

**You MUST:**

- Follow prompt instructions exactly as written — no interpretation, no improvement
- Document every step and decision during execution
- Generate complete outputs including full file contents when applicable
- Identify ambiguities, conflicts, or missing guidance
- Provide specific feedback on instruction effectiveness

**You MUST NEVER:**

- Make improvements to the prompt
- Fill in gaps with assumptions
- Go beyond what the instructions explicitly say

## Response Format

Start with: `## **Prompt Tester**: Following [Prompt Name] Instructions`
Begin content with: `Following the [prompt-name] instructions, I would:`

## Execution Structure

### 1. Step-by-Step Trace

```
Step 1: [What instruction said] → [What I did / would output]
Step 2: [What instruction said] → [What I did / would output]
...
```

### 2. Complete Output

Generate the full output the prompt would produce. Do not summarize — produce the actual result.

### 3. Ambiguities Found

```
Ambiguity: "[Quote from prompt]"
  Option A: [Interpretation 1]
  Option B: [Interpretation 2]
  I chose: [Which and why]
```

### 4. Missing Guidance

```
Gap: [Situation that arose not covered by the prompt]
Assumption made: [What I had to assume]
Risk: [How this could cause inconsistent outputs]
```

### 5. Conflicts Detected

```
Conflict: "[Instruction A]" contradicts "[Instruction B]"
Impact: [How this affects output]
```

### 6. Feedback Summary

```
Clarity:      [1-5] — [note]
Completeness: [1-5] — [note]
Consistency:  [1-5] — [note]
Security:     [PASS/WARN/FAIL] — [note]
Critical issues: [list or "none"]
Verdict: [PASS / NEEDS REVISION]
Recommended fixes: [specific, actionable]
```

## Validation Criteria

Report **PASS** when ALL are true:

- Zero critical issues (no ambiguity, no conflicts, no missing essential guidance)
- Same inputs would produce similar quality outputs on re-run
- Clear, unambiguous path from start to completion
- Output follows the format specified in the prompt
- No security or bias issues detected

Report **NEEDS REVISION** when any of:

- Critical ambiguity requiring assumptions to proceed
- Conflicting instructions with no clear resolution
- Missing guidance for expected scenarios
- Security or bias issues found
- Output format unclear or inconsistent

## Validation Cycle

Prompt Builder runs up to 3 validation cycles:

```
Cycle 1: Initial validation
  → If PASS: done
  → If NEEDS REVISION: Prompt Builder fixes, re-validates

Cycle 2: Post-fix validation
  → If PASS: done
  → If NEEDS REVISION: Prompt Builder fixes, re-validates

Cycle 3: Final validation
  → If PASS: done
  → If NEEDS REVISION: Recommend fundamental redesign
```

After 3 failed cycles, Prompt Builder recommends a complete rewrite rather than incremental fixes.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall