用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/carrot-foundation/methodology-rules --skill rule-error-handling命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | rule-error-handling |
| description | Rule mapping for error-handling |
Apply this rule whenever work touches:
*.tsProper error handling ensures that failures are visible, diagnosable, and recoverable. In a Lambda-based architecture, error propagation also controls retry behavior and dead-letter queue routing.
Validate external data where it enters the system. Use Zod schemas:
const parseResult = RuleInputSchema.safeParse(event);
if (!parseResult.success) {
logger.error({ errors: parseResult.error.issues }, 'Invalid rule input');
throw new ValidationError('Rule input failed schema validation', {
cause: parseResult.error,
});
}
const ruleInput = parseResult.data;
Internal function-to-function calls within a trusted boundary do not need redundant validation; rely on TypeScript's type system there.
When catching an error to add context, preserve the original error as the cause:
try {
await fetchDocument(documentId);
} catch (error) {
throw new DocumentFetchError(
`Failed to fetch document ${documentId} during credit evaluation`,
{ cause: error },
);
}
This preserves the full error chain for debugging while adding the business context needed to understand what was happening.
Lambda handlers must not swallow errors for operations that should be retried. Let the error propagate to the Lambda runtime:
// Correct - error reaches Lambda runtime, triggers retry
export const handler = async (event: SQSEvent): Promise<void> => {
const input = parseAndValidate(event);
await processRule(input);
};
// Wrong - error is caught and swallowed, message is lost
export const handler = async (event: SQSEvent): Promise<void> => {
try {
const input = parseAndValidate(event);
await processRule(input);
} catch {
console.log('Something went wrong');
}
};
Use pino for all logging. Include structured fields that aid debugging:
logger.error(
{
documentId,
ruleId,
operation: 'evaluateResult',
err: error,
},
'Rule evaluation failed',
);
Never include credentials, tokens, full request/response bodies with PII, or other sensitive data in log output.
For failure modes that callers are expected to handle (e.g., a document that legitimately fails validation), consider using a Result-like return type instead of throwing:
type EvaluationResult =
| { success: true; output: RuleOutput }
| { success: false; reason: string };
Reserve thrown errors for unexpected failures (infrastructure errors, programming bugs, corrupted data).