The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains a hardcoded credential in the SKILL.md file within an example authentication fixture.
Evidence: await page.getByLabel("Password").fill("TestPassword123!"); in SKILL.md.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and extract data from external web applications, creating a vulnerability surface where malicious content on a tested page could influence agent behavior.
Ingestion points: Data enters the context via page.goto(), page.waitForResponse(), and locators like usersPage.getUserRow() which process HTML content and API payloads.
Boundary markers: No explicit instructions or delimiters are provided to the agent to treat page content as untrusted data.
Capability inventory: The skill utilizes Bash (via npx playwright), and has Read, Edit, and Write file permissions.
Sanitization: There is no evidence of sanitization or filtering for the external content processed by the tests.
[COMMAND_EXECUTION]: The skill frequently invokes shell commands to install browsers and execute tests, which is expected for its purpose but increases the impact if the agent is compromised via injection.
Evidence: npx playwright install, npx playwright test, and docker compose up in SKILL.md and scripts/run-e2e-with-report.sh.

e2e-testing