메뉴
Enforces a strict input boundary protocol (detect, classify, filter, verify) to ensure untrusted data never reaches business logic raw.
기본적으로 소스를 먼저 확인하는 Prompt가 선택됩니다. 직접 명령으로 전환하거나 로컬 사본을 다운로드할 수도 있습니다.
설치 여부를 결정하기 전에 SKILL.md와 SkillsMP에 표시된 보조 파일을 읽어 보세요.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
직접 명령은 검토 Prompt를 거치지 않습니다. 실행하기 전에 소스를 확인하세요.
npx skills add https://github.com/ranbot-ai/awesome-skills --skill infinity명령은 한 줄로 유지됩니다. 복사하기 전에 가로로 스크롤해 전체 내용을 확인하세요.
로컬 사본을 원하시나요? SkillsMP에서 현재 제공할 수 있는 파일을 다운로드하세요.
SOC 직업 분류 기준
SKILL.md 표시 중
| name | infinity |
| description | Enforces a strict input boundary protocol (detect, classify, filter, verify) to ensure untrusted data never reaches business logic raw. |
| category | AI & Agents |
| source | antigravity |
| tags | ["api","ai","security","rag","cro"] |
| url | https://github.com/sickn33/antigravity-awesome-skills/tree/main/skills/infinity |
Nothing untrusted ever reaches the core — it is stopped before contact. No external data touches the codebase raw. Every boundary where data enters the system must have a filter.
The #1 source of silent bugs, crashes, and vulnerabilities is external data that arrives in an unexpected shape and gets used directly without checking. This skill enforces a filter layer at every entry point, every time.
.body, .params, .query, .env, fs.read, or a third-party SDK responseBefore writing or modifying any code that involves external data, the AI must identify and list every entry point in scope:
The AI must not write any data-handling logic until every entry point in scope is listed.
For every entry point identified, the AI classifies it into one of three trust levels:
| Level | Definition | Examples |
|---|---|---|
TRUSTED | Internal constants, hardcoded values, your own compile-time config | Enum values, hardcoded defaults, internal constants |
SEMI-TRUSTED | Your own internal services, internal APIs, controlled infrastructure | Internal microservice responses, your own database reads |
UNTRUSTED | Anything from users, the internet, third parties, or the filesystem | User input, external API responses, uploaded files, env vars, CLI args |
Rule:
TRUSTEDinputs may be used directly.SEMI-TRUSTEDandUNTRUSTEDinputs must pass through a filter layer before any use.
The AI outputs this classification before writing any handling code:
INFINITY — BOUNDARY MAP
─────────────────────────────────────────
Entry Point | Trust Level | Filter Required
─────────────────────────────────────────
req.body.email | UNTRUSTED | ✓ format + sanitize
process.env.API_KEY | UNTRUSTED | ✓ presence + non-empty
internalService.getData()| SEMI-TRUSTED | ✓ schema validate
PAGINATION_LIMIT = 20 | TRUSTED | ✗ none needed
─────────────────────────────────────────
Every UNTRUSTED and SEMI-TRUSTED input must pass through validation before it reaches any business logic, storage, or rendering. The AI must apply the right filter type for the right context:
Type Checking
Schema Validation
Sanitization
Presence & Format Checks
Rejection Rule
// WRONG — using raw input directly
const user = await db.find(req.params.id);
// RIGHT — validate before use
const id = req.params.id;
if (!id || typeof id !== 'string' || !isValidUUID(id)) {
return res.status(400).json({ error: 'Invalid ID format' });
}
const user = await db.find(id);
Before the AI declares any data-handling code complete, it traces each entry point and confirms:
INFINITY — VERIFICATION
─────────────────────────────────────────
Entry Point | Filter Exists | Filter Type
─────────────────────────────────────────
req.body.email | ✓ YES | format + sanitize
process.env.API_KEY | ✓ YES | presence check
internalService.getData()| ✓ YES | schema validation
─────────────────────────────────────────
Unfiltered inputs reaching logic: NONE ✓
─────────────────────────────────────────
If any UNTRUSTED or SEMI-TRUSTED input reaches logic, storage, or rendering without a filter — the AI flags it. It does not silently pass.