Exposed secret literal
- Finding
- File appears to expose a hardcoded API secret or token.
Security checks across static analysis, malware telemetry, and agentic risk
This is a plausible RPA automation skill, but it needs careful review because it stores/reuses login cookies and API keys, can replay broad browser/API actions, and weakens TLS/browser safety defaults.
Install only if you understand and accept the credential and automation risks. Before replaying tasks, inspect the generated Python, avoid storing real secrets in scripts, protect or delete saved cookie files, keep TLS verification and browser sandboxing enabled where possible, and require manual confirmation for purchases, POST requests, or other account-changing actions.
1/46 vendors flagged this skill as malicious, and 45/46 flagged it as clean.
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A generated RPA script may contain live API keys and can keep using those keys outside the original chat session.
The skill documents writing real API secrets into persistent generated Python scripts, increasing the chance of key leakage through files, backups, sharing, or git commits.
代码生成器检测到 `env` 字段后,会把真实密钥**直接写入生成脚本**——回放时**无需 `export`**,脚本可直接运行。
Prefer environment variables or a secret manager, use least-privilege API keys, never commit generated scripts containing secrets, and rotate keys after testing.
Anyone or any script with access to those cookie files may be able to act as the logged-in user without re-entering a password, OTP, CAPTCHA, or QR login.
The skill persists authenticated session cookies on disk and automatically reuses them for future recordings and replays.
`#rpa-login-done` → 自动导出 Cookie,保存到 `~/.openclaw/rpa/sessions/passport.ctrip.com/cookies.json` ... 录制/回放时自动注入 Cookie
Use this only with accounts you are allowed to automate, protect the session directory with strict file permissions, avoid sensitive accounts, and delete or refresh saved cookies when no longer needed.
A replayed script could submit forms, modify third-party data, trigger purchases, or call authenticated APIs exactly as recorded.
The skill exposes broad raw API mutation capability and direct replay of saved automations without artifact-shown per-run approval for high-impact actions.
**HTTP API** | `GET` / `POST` any REST endpoint ... `#rpa-run:{task name}` ... run one directlyInspect generated scripts before replay, separate read-only from write actions, require manual confirmation for POST/submit/checkout steps, and use test accounts where possible.
Network attackers or misconfigured proxies could intercept or alter API responses, and API keys sent through those requests may be exposed.
HTTPS certificate verification is disabled for an HTTP client in code associated with the recorder/API flow.
async with httpx.AsyncClient(timeout=_t, verify=False) as hc:
Enable TLS verification by default, remove verify=False, and require an explicit user-controlled opt-out only for local testing.
This may bypass site automation controls or violate site terms, and no-sandbox increases the local impact of a malicious webpage opened by the automation.
Generated scripts hide Playwright automation signals and disable Chromium sandboxing by default.
args=["--no-sandbox", "--disable-blink-features=AutomationControlled"] ... `Object.defineProperty(navigator, 'webdriver', {get: () => undefined})`Do not use this on sites that forbid automation, remove stealth settings unless explicitly needed and permitted, and keep Chromium sandboxing enabled for normal desktop use.