Skills test-engineer Audit History
📦

Audit History

test-engineer - 4 audits

Version comparison

Capability and finding changes across audited versions, newest first.

VersionDateResultReview itemsChange vs previous
v4 LatestJul 24, 2026, 12:12 AM 1 confirmed0No capability change
v3 Jul 15, 2026, 03:52 PM 1 confirmed8No capability change
v2 Jul 15, 2026, 03:52 PM 1 confirmed8No capability change
v1 Jul 15, 2026, 03:52 PM 1 confirmed8Baseline

Jul 24, 2026, 12:12 AM

All 19 static findings are false positives caused by Markdown formatting and multilingual prose. The skill still directs active API and load testing without authorization or production safeguards. Add explicit consent, scope, rate, and environment checks before publication.

1
Files scanned
216
Lines analyzed
2
Review items
0
False positives ignored

Confirmed security concerns (1)

High
Active Testing Lacks Authorization Guardrails
The skill directs real API calls and performance load tests without requiring target ownership, production approval, scope, or traffic limits.
The workflow explicitly requires actual API and load-test execution. No authorization, environment, or rate-limit check appears in that workflow.
Audited by: codex

Jul 15, 2026, 03:52 PM

Eight static findings are confirmed because their cited ranges direct the agent to install dependencies or execute test, load, Python, and container commands. Eleven findings are false positives caused by Markdown backticks, code fences, tool names, or UTF-8 Chinese documentation rather than executable or obfuscated content. Active API and load testing also lacks explicit authorization and environment safeguards.

1
Files scanned
216
Lines analyzed
10
Review items
0
False positives ignored

Confirmed security concerns (1)

High
Missing Authorization Safeguards for Active Testing
The skill requires real API calls and load tests but does not require target ownership, explicit authorization, non-production confirmation, traffic limits, or stop conditions.
The cited workflows explicitly require actual API and performance test execution. The document contains no authorization or production-safety gate for those actions.
Capability review items (8)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
Ruby/shell backtick execution · 3 occurrences
```
The cited example directs the agent to run pip install and pytest. Installing project requirements and executing tests can run untrusted package or repository code without an approval boundary.
Medium
Ruby/shell backtick execution
| Web UI自动化 | Playwright | `pip install playwright && playwright install` → 生成脚本 → `pytest` 运行 |
Line 172 explicitly chains pip install, Playwright installation, and pytest execution. Dependency installation and test execution modify the environment and can execute untrusted code.
Medium
Ruby/shell backtick execution
| API测试 | Python requests / curl | 生成脚本 → `python test_api.py` 执行 |
The range instructs the agent to execute a generated Python API test and includes load-test commands. Generated scripts and supplied targets are not constrained by approval or isolation.
Medium
Ruby/shell backtick execution
| 性能测试 | k6 / wrk / hey | `k6 run script.js` 或 `wrk -t4 -c100 -d30s URL` |
Line 174 provides direct k6 and wrk command invocations against a URL. These commands generate substantial traffic and can disrupt an unintended target.
Medium
Ruby/shell backtick execution
| 单元测试 | pytest / JUnit / Jest | `pytest -v --tb=short --junitxml=report.xml` |
The cited range includes direct pytest execution and a Docker Compose command. Both execute project-controlled configuration or code and can change the local environment.
Medium
Ruby/shell backtick execution
| 容器化 | Docker Compose | `docker compose up -d` 启动测试环境 |
Line 179 directs the agent to start services with Docker Compose. Repository-controlled compose files can run images, expose ports, mount files, and alter the host environment.
Audited by: codex

Jul 15, 2026, 03:52 PM

Eight static findings are confirmed because their cited ranges direct the agent to install dependencies or execute test, load, Python, and container commands. Eleven findings are false positives caused by Markdown backticks, code fences, tool names, or UTF-8 Chinese documentation rather than executable or obfuscated content. Active API and load testing also lacks explicit authorization and environment safeguards.

1
Files scanned
216
Lines analyzed
10
Review items
0
False positives ignored

Confirmed security concerns (1)

High
Missing Authorization Safeguards for Active Testing
The skill requires real API calls and load tests but does not require target ownership, explicit authorization, non-production confirmation, traffic limits, or stop conditions.
The cited workflows explicitly require actual API and performance test execution. The document contains no authorization or production-safety gate for those actions.
Capability review items (8)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
Ruby/shell backtick execution · 3 occurrences
```
The cited example directs the agent to run pip install and pytest. Installing project requirements and executing tests can run untrusted package or repository code without an approval boundary.
Medium
Ruby/shell backtick execution
| Web UI自动化 | Playwright | `pip install playwright && playwright install` → 生成脚本 → `pytest` 运行 |
Line 172 explicitly chains pip install, Playwright installation, and pytest execution. Dependency installation and test execution modify the environment and can execute untrusted code.
Medium
Ruby/shell backtick execution
| API测试 | Python requests / curl | 生成脚本 → `python test_api.py` 执行 |
The range instructs the agent to execute a generated Python API test and includes load-test commands. Generated scripts and supplied targets are not constrained by approval or isolation.
Medium
Ruby/shell backtick execution
| 性能测试 | k6 / wrk / hey | `k6 run script.js` 或 `wrk -t4 -c100 -d30s URL` |
Line 174 provides direct k6 and wrk command invocations against a URL. These commands generate substantial traffic and can disrupt an unintended target.
Medium
Ruby/shell backtick execution
| 单元测试 | pytest / JUnit / Jest | `pytest -v --tb=short --junitxml=report.xml` |
The cited range includes direct pytest execution and a Docker Compose command. Both execute project-controlled configuration or code and can change the local environment.
Medium
Ruby/shell backtick execution
| 容器化 | Docker Compose | `docker compose up -d` 启动测试环境 |
Line 179 directs the agent to start services with Docker Compose. Repository-controlled compose files can run images, expose ports, mount files, and alter the host environment.
Audited by: codex

Jul 15, 2026, 03:52 PM

Eight static findings are confirmed because their cited ranges direct the agent to install dependencies or execute test, load, Python, and container commands. Eleven findings are false positives caused by Markdown backticks, code fences, tool names, or UTF-8 Chinese documentation rather than executable or obfuscated content. Active API and load testing also lacks explicit authorization and environment safeguards.

1
Files scanned
216
Lines analyzed
10
Review items
0
False positives ignored

Confirmed security concerns (1)

High
Missing Authorization Safeguards for Active Testing
The skill requires real API calls and load tests but does not require target ownership, explicit authorization, non-production confirmation, traffic limits, or stop conditions.
The cited workflows explicitly require actual API and performance test execution. The document contains no authorization or production-safety gate for those actions.
Capability review items (8)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

Medium
Ruby/shell backtick execution · 3 occurrences
```
The cited example directs the agent to run pip install and pytest. Installing project requirements and executing tests can run untrusted package or repository code without an approval boundary.
Medium
Ruby/shell backtick execution
| Web UI自动化 | Playwright | `pip install playwright && playwright install` → 生成脚本 → `pytest` 运行 |
Line 172 explicitly chains pip install, Playwright installation, and pytest execution. Dependency installation and test execution modify the environment and can execute untrusted code.
Medium
Ruby/shell backtick execution
| API测试 | Python requests / curl | 生成脚本 → `python test_api.py` 执行 |
The range instructs the agent to execute a generated Python API test and includes load-test commands. Generated scripts and supplied targets are not constrained by approval or isolation.
Medium
Ruby/shell backtick execution
| 性能测试 | k6 / wrk / hey | `k6 run script.js` 或 `wrk -t4 -c100 -d30s URL` |
Line 174 provides direct k6 and wrk command invocations against a URL. These commands generate substantial traffic and can disrupt an unintended target.
Medium
Ruby/shell backtick execution
| 单元测试 | pytest / JUnit / Jest | `pytest -v --tb=short --junitxml=report.xml` |
The cited range includes direct pytest execution and a Docker Compose command. Both execute project-controlled configuration or code and can change the local environment.
Medium
Ruby/shell backtick execution
| 容器化 | Docker Compose | `docker compose up -d` 启动测试环境 |
Line 179 directs the agent to start services with Docker Compose. Repository-controlled compose files can run images, expose ports, mount files, and alter the host environment.
Audited by: codex