The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses a sensitive authentication state file to facilitate automated login in tests.\n
Evidence: Both SKILL.md and references/acceptance-spec-template.md reference the file path e2e/.auth/user.json. This file is standard in Playwright for storing session cookies and authentication tokens, making it a high-value target for exposure.\n- [COMMAND_EXECUTION]: The workflow involves executing shell commands to run tests that were generated during the same session.\n
Evidence: SKILL.md specifies the command npx playwright test --project="chromium:acceptance" {TICKET}-{description}.spec.ts. This executes code created by the agent onto the local system.\n- [PROMPT_INJECTION]: The skill processes untrusted input from user requests (ticket keys and descriptions) to generate both filenames and test logic, creating a surface for indirect prompt injection.\n
Ingestion points: User-provided [TICKET] and [description] arguments in SKILL.md.\n
Boundary markers: Absent; there are no instructions or delimiters to isolate the ticket data from the rest of the generated test code.\n
Capability inventory: The skill can write files to the e2e/ directory and execute shell commands through npx.\n
Sanitization: Absent; the skill does not define any validation or escaping mechanisms for the ticket data before it is interpolated into shell commands or TypeScript files.

create-acceptance-test