Security audit
Security checks for vulnerabilities and agentic risk
The skill is a coherent scenario-testing assistant, but it grants broad live-app interaction and repository mutation authority without enough scoping or user control.
Install only if you are comfortable with a skill that may crawl live applications, use provided authentication, exercise UI/API flows, and modify plus commit repository code. Prefer running it against staging or test data, require an explicit target and read-only mode first, approve any mutation testing, external report sharing, code edits, and commits manually.
The skill states that destructive actions are only simulated, but the documented workflow performs real UI interaction and runtime verification against live targets. In practice, clicking delete/save/submit paths or exercising state-changing flows can trigger actual mutations, causing unintended data loss or unauthorized changes despite the safety claim.
Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).
| GET /items/:id | 단일 항목 객체 반환 |
| POST /items | 생성된 항목 반환 + 201 |
| PUT /items/:id | 수정된 항목 반환 + 200 |
| DELETE /items/:id | 204 또는 200 + 이후 GET 시 404 |
| POST /auth/login | 토큰/세션 반환 |
| POST /auth/logout | 세션 무효화 + 이후 인증 API 401 |
The skill goes beyond analysis by automatically modifying code, running tests, applying fixes, and committing changes. This expands the trust boundary from detection to autonomous repository mutation, which can introduce incorrect changes, persistence of bad fixes, or unauthorized edits if the target or inferred defect is wrong.
The trigger phrase "SDD 돌려줘" is overly broad and can activate the skill from ordinary conversation without clearly identifying a target or scope. Ambiguous activation increases the chance of unintended crawling, code analysis, or attempted modification against the wrong resource.
The phrase "이 프로젝트 시나리오 검증해" does not define project type, location, authorization, or whether analysis should be read-only or mutating. That ambiguity can lead the skill to operate on unintended codebases or assume permissions that were never explicitly granted.
The trigger "화면 흐름 논리적으로 맞는지 확인해" is generic enough to overlap with routine review or discussion requests. In a skill capable of crawling sites, executing flows, generating reports, and proposing fixes, vague activation language materially raises the risk of unintended actions.
The skill instructs saving reports to files and delivering them through external communication channels such as Discord or Slack. Security and logic-test reports often contain sensitive URLs, code paths, defect details, and inferred system behavior, so automatic external distribution can leak internal information beyond the intended environment.
Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.
copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
This markdown file contains user-facing natural-language guidance exclusively in Korean, while the policy forbids forcing a specific language without user opt-in. There is no indication that the skill is region-specific or that users can choose another language.
No suspicious patterns detected.