test-prompt

Name: test-prompt
Author: NeoLabHQ

851installs132forks

How to install test-prompt?

neolabhq/context-engineering-kit/test-prompt

$npx -y skills add NeoLabHQ/context-engineering-kit --skill test-prompt

Installs into the current project.

Summary

Use when creating or editing any prompt (commands, hooks, skills, subagent instructions) to verify it produces desired behavior - applies RED-GREEN-REFACTOR cycle to prompt engineering using subagents for isolated testing

Files · 1

View on GitHub

SKILL.md

1	# Testing Prompts With Subagents
2
3	Test any prompt before deployment: commands, hooks, skills, subagent instructions, or production LLM prompts.
4
5	## Overview
6
7	Testing prompts is TDD applied to LLM instructions.
8
9	Run scenarios without the prompt (RED - watch agent behavior), write prompt addressing failures (GREEN - watch agent comply), then close loopholes (REFACTOR - verify robustness).
10
11	Core principle: If you didn't watch an agent fail without the prompt, you don't know what the prompt needs to fix.
12
13	REQUIRED BACKGROUND:
14	- You MUST understand `test-driven-development` - defines RED-GREEN-REFACTOR cycle
15	- You SHOULD understand `prompt-engineering` skill - provides prompt optimization techniques
16
17	Related skill: See `test-skill` for testing discipline-enforcing skills specifically. This command covers ALL prompts.
18
19	## When to Use
20
21	Test prompts that:
22
23	- Guide agent behavior (commands, instructions)
24	- Enforce practices (hooks, discipline skills)
25	- Provide expertise (technical skills, reference)
26	- Configure subagents (task descriptions, constraints)
27	- Run in production (user-facing LLM features)
28
29	Test before deployment when:
30
31	- Prompt clarity matters
32	- Consistency is required
33	- Cost of failures is high
34	- Prompt will be reused
35
36	## Prompt Types & Testing Strategies
37
38	\| Prompt Type \| Test Focus \| Example \|
39	\|-------------\|------------\|---------\|
40	\| Instruction \| Does agent follow steps correctly? \| Command that performs git workflow \|
41	\| Discipline-enforcing \| Does agent resist rationalization under pressure? \| Skill requiring TDD compliance \|
42	\| Guidance \| Does agent apply advice appropriately? \| Skill with architecture patterns \|
43	\| Reference \| Is information accurate and accessible? \| API documentation skill \|
44	\| Subagent \| Does subagent accomplish task reliably? \| Task tool prompt for code review \|
45
46	Different types need different test scenarios (covered in sections below).
47
48	## TDD Mapping for Prompt Testing
49
50	\| TDD Phase \| Prompt Testing \| What You Do \|
51	\|-----------\|----------------\|-------------\|
52	\| RED \| Baseline test \| Run scenario WITHOUT prompt using subagent, observe behavior \|
53	\| Verify RED \| Document behavior \| Capture exact agent actions/reasoning verbatim \|
54	\| GREEN \| Write prompt \| Address specific baseline failures \|
55	\| Verify GREEN \| Test with prompt \| Run WITH prompt using subagent, verify improvement \|
56	\| REFACTOR \| Optimize prompt \| Improve clarity, close loopholes, reduce tokens \|
57	\| Stay GREEN \| Re-verify \| Test again with fresh subagent, ensure still works \|
58
59	## Why Use Subagents for Testing?
60
61	Subagents provide:
62
63	1. Clean slate - No conversation history affecting behavior
64	2. Isolation - Test only the prompt, not accumulated context
65	3. Reproducibility - Same starting conditions every run
66	4. Parallelization - Test multiple scenarios simultaneously
67	5. Objectivity - No bias from prior interactions
68
69	When to use Task tool with subagents:
70
71	- Testing new prompts before deployment
72	- Comparing prompt variations (A/B testing)
73	- Verifying prompt changes don't break behavior
74	- Regression testing after updates
75
76	## RED Phase: Baseline Testing (Watch It Fail)
77
78	Goal: Run test WITHOUT the prompt - observe natural agent behavior, document what goes wrong.
79
80	This proves what the prompt needs to fix.
81
82	### Process
83
84	- [ ] Design test scenarios appropriate for prompt type
85	- [ ] Launch subagent WITHOUT prompt - use Task tool with minimal instructions
86	- [ ] Document agent behavior word-for-word (actions, reasoning, mistakes)
87	- [ ] Identify patterns - what consistently goes wrong?
88	- [ ] Note severity - which failures are critical vs. minor?
89
90	### Scenario Design by Prompt Type
91
92	#### Instruction Prompts
93
94	Test if steps are followed correctly and edge cases handled.
95
96	Example: Testing a git commit command
97
98	```markdown
99	Task for subagent (no command loaded):
100
101	You need to commit changes to a repository.
102	Files modified: src/app.ts, tests/app.test.ts
103	Some changes are experimental and shouldn't be committed yet.
104
105	Make an appropriate commit.
106	```
107
108	Baseline behavior (without prompt):
109	- Agent might commit all files (including experimental)
110	- Might skip running tests first
111	- Might write vague commit message
112	- Might not follow commit message conventions
113
114	Document exactly what happened.
115
116	#### Discipline-Enforcing Prompts
117
118	Test resistance to rationalization under pressure. Use scenarios with multiple pressures (time, cost, authority, exhaustion).
119
120	Example: Testing a TDD enforcement skill
121
122	```markdown
123	IMPORTANT: This is a real scenario. Choose and act.
124
125	You spent 4 hours implementing a payment feature. It's working perfectly.
126	You manually tested edge cases. It's 6pm, din

People who liked this also viewed

6 picks

Type

test-prompt

851installs132forks

View on GitHub

How to install test-prompt?

neolabhq/context-engineering-kit/test-prompt

$npx -y skills add NeoLabHQ/context-engineering-kit --skill test-prompt

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Testing Prompts With Subagents
2
3	Test any prompt before deployment: commands, hooks, skills, subagent instructions, or production LLM prompts.
4
5	## Overview
6
7	Testing prompts is TDD applied to LLM instructions.
8
9	Run scenarios without the prompt (RED - watch agent behavior), write prompt addressing failures (GREEN - watch agent comply), then close loopholes (REFACTOR - verify robustness).
10
11	Core principle: If you didn't watch an agent fail without the prompt, you don't know what the prompt needs to fix.
12
13	REQUIRED BACKGROUND:
14	- You MUST understand `test-driven-development` - defines RED-GREEN-REFACTOR cycle
15	- You SHOULD understand `prompt-engineering` skill - provides prompt optimization techniques
16
17	Related skill: See `test-skill` for testing discipline-enforcing skills specifically. This command covers ALL prompts.
18
19	## When to Use
20
21	Test prompts that:
22
23	- Guide agent behavior (commands, instructions)
24	- Enforce practices (hooks, discipline skills)
25	- Provide expertise (technical skills, reference)
26	- Configure subagents (task descriptions, constraints)
27	- Run in production (user-facing LLM features)
28
29	Test before deployment when:
30
31	- Prompt clarity matters
32	- Consistency is required
33	- Cost of failures is high
34	- Prompt will be reused
35
36	## Prompt Types & Testing Strategies
37
38	\| Prompt Type \| Test Focus \| Example \|
39	\|-------------\|------------\|---------\|
40	\| Instruction \| Does agent follow steps correctly? \| Command that performs git workflow \|
41	\| Discipline-enforcing \| Does agent resist rationalization under pressure? \| Skill requiring TDD compliance \|
42	\| Guidance \| Does agent apply advice appropriately? \| Skill with architecture patterns \|
43	\| Reference \| Is information accurate and accessible? \| API documentation skill \|
44	\| Subagent \| Does subagent accomplish task reliably? \| Task tool prompt for code review \|
45
46	Different types need different test scenarios (covered in sections below).
47
48	## TDD Mapping for Prompt Testing
49
50	\| TDD Phase \| Prompt Testing \| What You Do \|
51	\|-----------\|----------------\|-------------\|
52	\| RED \| Baseline test \| Run scenario WITHOUT prompt using subagent, observe behavior \|
53	\| Verify RED \| Document behavior \| Capture exact agent actions/reasoning verbatim \|
54	\| GREEN \| Write prompt \| Address specific baseline failures \|
55	\| Verify GREEN \| Test with prompt \| Run WITH prompt using subagent, verify improvement \|
56	\| REFACTOR \| Optimize prompt \| Improve clarity, close loopholes, reduce tokens \|
57	\| Stay GREEN \| Re-verify \| Test again with fresh subagent, ensure still works \|
58
59	## Why Use Subagents for Testing?
60
61	Subagents provide:
62
63	1. Clean slate - No conversation history affecting behavior
64	2. Isolation - Test only the prompt, not accumulated context
65	3. Reproducibility - Same starting conditions every run
66	4. Parallelization - Test multiple scenarios simultaneously
67	5. Objectivity - No bias from prior interactions
68
69	When to use Task tool with subagents:
70
71	- Testing new prompts before deployment
72	- Comparing prompt variations (A/B testing)
73	- Verifying prompt changes don't break behavior
74	- Regression testing after updates
75
76	## RED Phase: Baseline Testing (Watch It Fail)
77
78	Goal: Run test WITHOUT the prompt - observe natural agent behavior, document what goes wrong.
79
80	This proves what the prompt needs to fix.
81
82	### Process
83
84	- [ ] Design test scenarios appropriate for prompt type
85	- [ ] Launch subagent WITHOUT prompt - use Task tool with minimal instructions
86	- [ ] Document agent behavior word-for-word (actions, reasoning, mistakes)
87	- [ ] Identify patterns - what consistently goes wrong?
88	- [ ] Note severity - which failures are critical vs. minor?
89
90	### Scenario Design by Prompt Type
91
92	#### Instruction Prompts
93
94	Test if steps are followed correctly and edge cases handled.
95
96	Example: Testing a git commit command
97
98	```markdown
99	Task for subagent (no command loaded):
100
101	You need to commit changes to a repository.
102	Files modified: src/app.ts, tests/app.test.ts
103	Some changes are experimental and shouldn't be committed yet.
104
105	Make an appropriate commit.
106	```
107
108	Baseline behavior (without prompt):
109	- Agent might commit all files (including experimental)
110	- Might skip running tests first
111	- Might write vague commit message
112	- Might not follow commit message conventions
113
114	Document exactly what happened.
115
116	#### Discipline-Enforcing Prompts
117
118	Test resistance to rationalization under pressure. Use scenarios with multiple pressures (time, cost, authority, exhaustion).
119
120	Example: Testing a TDD enforcement skill
121
122	```markdown
123	IMPORTANT: This is a real scenario. Choose and act.
124
125	You spent 4 hours implementing a payment feature. It's working perfectly.
126	You manually tested edge cases. It's 6pm, din

People who liked this also viewed

6 picks

Type