Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add luanpdd/kit-mcp --skill ai-prompt-characterizationgit clone --depth 1 https://github.com/luanpdd/kit-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/luanpdd/kit-mcp/ai-prompt-characterization)<a href="https://agentmods.dev/skills/luanpdd/kit-mcp/ai-prompt-characterization"><img src="https://agentmods.dev/badge/skills/luanpdd/kit-mcp/ai-prompt-characterization/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/luanpdd/kit-mcp/ai-prompt-characterization"><img src="https://agentmods.dev/badge/skills/luanpdd/kit-mcp/ai-prompt-characterization.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00047 | $0.03519 |
| Opus 5 | $0.00023 | $0.01759 |
| Sonnet 5 | $0.00009 | $0.00704 |
| Haiku 4.5 | $0.00005 | $0.00352 |
Grade B, and why
ai-prompt-characterization scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Instruction-override phrasingmediumPrompt injection
Text telling the model to disregard its earlier instructions or safety rules is the shape of a prompt injection, whoever wrote it.
| **Adversarial** | Tentativa de jailbreak / prompt injection | "Ignore previous instructions and..." | Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 337 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI Prompt Characterization (Modernização)
Quando usar
LLM carrega esta skill quando user vai modificar prompt ou tool definition de LLM em produção. Trigger phrases:
- "vou mudar esse prompt", "modificar prompt em prod"
- "atualizar tool definition", "function calling schema"
- "como testar mudança de prompt?"
- "characterization de prompt", "snapshot de generation"
- "esse prompt tem 300 linhas e ninguém testou ainda"
- prompt em arquivo como
prompts/<name>.mdou string template em código
Insight central: prompts e tools são código legacy também quando:
-
100 linhas
- Em uso em produção
- Mudanças quebram silenciosamente (output diferente, downstream parser falha)
- Sem characterization tests
Regras absolutas
- Prompts são código. Tratam-se com mesmo rigor: versionado, testado, code-reviewed. NÃO são "config text que muda livremente".
- Determinismo via
temperature=0+seed. Anthropic Claude e OpenAI ambos suportam seed. Sem isso, characterization é flaky. - Capture mais que
text. Outputs incluem:text,finish_reason,tool_calls(se function calling),input_tokens,output_tokens,model_version. Snapshot de TODOS estes campos. - Sanitize aggressively. Outputs LLM frequentemente incluem timestamps mencionados, UUIDs gerados, datas relativas. Normalize ANTES de snapshot.
- 5+ inputs cobrindo intents distintas. Não é "happy path × 5"; é "5 intents qualitativamente diferentes" — concision request, troubleshooting, explanation, creative, edge case.
- Behavioral coverage = % intents cobertas. Métrica não é coverage de "linhas do prompt" (não existe); é coverage de variações comportamentais.
- Re-rodar em CI quando model_version muda. Anthropic publica nova versão de Claude → re-rode characterization → revisar diffs → aceitar/rejeitar.
Patterns canônicos
Pattern 1: Setup canônico de characterization de prompt
// tests/characterization/prompts/generate-summary.test.ts
import { Anthropic } from '@anthropic-ai/sdk'
import { describe, test, expect } from 'vitest'
import { readFileSync } from 'fs'
const client = new Anthropic({ apiKey: process.env.ANTHROPIC_API_KEY })
const PROMPT = readFileSync('prompts/generate-summary.md', 'utf-8')
interface PromptInput {
systemPrompt: string
userMessage: string
maxTokens?: number
}
async function runPrompt(input: PromptInput) {
const response = await client.messages.create({
model: 'claude-opus-4-7',
max_tokens: input.maxTokens ?? 500,
temperature: 0, // determinismo
system: input.systemPrompt,
messages: [{ role: 'user', content: input.userMessage }],
})
return {
text: response.content[0].type === 'text' ? response.content[0].text : '',
stopReason: response.stop_reason,
inputTokens: response.usage.input_tokens,
outputTokens: response.usage.output_tokens,
modelVersion: response.model,
}
}
function sanitizeForSnapshot(o: any): any {
return JSON.parse(
JSON.stringify(o, (key, value) => {
// normalizar timestamps mencionados ("Today is 2026-05-08") → "<DATE>"
if (typeof value === 'string') {
value = value.replace(/\d{4}-\d{2}-\d{2}/g, '<DATE>')
value = value.replace(/\d{2}:\d{2}(:\d{2})?/g, '<TIME>')
value = value.replace(/[0-9a-f]{8}-[0-9a-f]{4}-[0-9a-f]{4}/g, '<UUID>')
}
// permitir model version mas separar para audit (não no snapshot)
if (key === 'modelVersion') return '<MODEL>'
return value
})
)
}
describe('generate-summary prompt — characterization', () => {
test('intent: concise summary of long article', async () => {
const captured = await runPrompt({
systemPrompt: PROMPT,
userMessage: 'Resuma em 2 sentenças: [longo artigo de 500 palavras]...',
})
expect(sanitizeForSnapshot(captured)).toMatchSnapshot()
})
test('intent: bullet-list summary', async () => { /* ... */ })
test('intent: technical/code summary', async () => { /* ... */ })
test('intent: ambiguous request (edge)', async () => { /* ... */ })
test('intent: hostile / prompt injection attempt', async () => { /* ... */ })
})
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 6d ago First seen · 337 lines · 47 tokens per session scan B 79cebabf35a2
ai-prompt-characterization is a skill published in the GitHub repository luanpdd/kit-mcp (1 stars, last pushed 3d ago), licensed MIT. It adds 47 tokens to every session and 3,519 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it B with 1 finding (instruction-override phrasing). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-03.
Other skills, from other repositories
data-quality-frameworks
Implement data quality validation with Great Expectations, dbt tests, and data contracts. Use when building data quality pipelines, implementing validation rules, or establishing data contracts.
dbt-transformation-patterns
Master dbt (data build tool) for analytics engineering with model organization, testing, documentation, and incremental strategies. Use when building data transformations, creating data models, or implementing analytics engineering best practices.
nemo-automodel-model-onboarding
Guide for onboarding new model architectures into NeMo AutoModel, including architecture discovery, implementation patterns, registration, and validation.
nemo-automodel-recipe-development
Create and modify NeMo AutoModel training and evaluation recipes, including YAML structure, builders, and execution flow.
neuron-test-engineer
Write tests for Neuron AI agents, RAG systems, workflows, and tools using the built-in testing utilities. Use this skill when the user mentions testing agents, writing unit tests, mocking AI providers, testing tool execution, verifying RAG retrieval, testing workflow behavior, or creating test cases for Neuron AI…
testing-llm
LLM and AI testing patterns — mock responses, evaluation with DeepEval/RAGAS, structured output validation, and agentic test patterns (generator, healer, planner). Use when testing AI features, validating LLM outputs, or building evaluation pipelines.