generate-web-skills

SUSPICIOUS: the skill's capabilities mostly match its stated purpose and its install path is from official tooling, so it does not look malicious. However, it captures and reuses live authenticated session data, stores it locally, and can generate shareable skills embedding those credentials, creating meaningful credential-handling and data-leak risk disproportionate to a normal documentation skill.

This module is not overtly self-propagating malware, but it is explicitly designed to extract and persist highly sensitive authentication secrets (host-scoped cookies and authorization/token/CSRF header values) from an interactive Playwright login session into a user-specified HAR file, while using a persistent browser profile on disk. If the package is untrusted or if outHar/PW_SETUP_PROFILE are attacker-influenced, it can readily facilitate credential theft or secret leakage; treat outputs and the profile directory as sensitive credentials and apply strict path/domain allowlisting and file access controls.

This file is best characterized as a powerful orchestration harness rather than inherently malicious code. The dominant risk is that it dynamically loads and executes a JavaScript module from a path supplied at runtime (arbitrary code execution with Playwright page authority). A secondary risk is optional injection of session cookies into a persistent browser context, which can enable authenticated actions performed by that loaded module. If variantPath/cookiesPath/outDir/PW_PROFILE/PW_PARAMS are strictly controlled and not attacker-influenced, risk is substantially reduced; otherwise, the module should be treated as high-impact and unsafe to use with untrusted inputs.

No direct malicious behavior is evidenced in the provided bash wrapper itself (no explicit credential theft, backdoor mechanics, or exfiltration). However, the script materially increases security exposure by (1) performing runtime npm/npx installs of third-party tooling and browser binaries without showing pinning, and (2) delegating all sensitive login/session capture behavior to an external script (setup_login.js) whose actions are not verifiable from this fragment. Additionally, it produces HAR files that likely contain authentication/session material, which is high-value data requiring careful handling. Overall: likely a legitimate setup tool, but with moderate security risk primarily due to supply-chain execution and high-sensitivity session artifact creation.

This bash wrapper itself shows no direct malicious behavior (no hardcoded secrets, no obvious exfiltration, no backdoor-like actions, and no obfuscated payloads). The primary security concerns are (1) automatic npm installation and Playwright asset downloads from external sources without evidence of pinning/integrity enforcement in this fragment, and (2) indirect execution of caller/lesson-controlled code through run_variant.js with a runtime-supplied VARIANT module. If lesson directories/variant modules or dependency download sources are not trusted and pinned, the risk increases to medium due to supply-chain and arbitrary automation/code-execution potential in the Node/Playwright execution context.