The Agent Skills Directory

[COMMAND_EXECUTION]: The skill workflow involves executing shell commands via npx playwright test using an interpolated {feature} name derived from user input. This creates a potential command injection surface if the feature name is not validated to exclude shell metacharacters.
Evidence: Found in SKILL.md workflow and checklist sections.
[DATA_EXFILTRATION]: Generated test code includes references to e2e/.auth/user.json for loading storageState. This file typically contains sensitive authentication session data such as cookies or bearer tokens.
Evidence: Found in SKILL.md and references/spec-template.md within the beforeAll block.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by incorporating untrusted user input into the code generation and execution flow.
Ingestion points: User-supplied feature-name argument documented in SKILL.md frontmatter.
Boundary markers: None identified; the input is used directly for file naming and shell command arguments.
Capability inventory: File system write operations in the e2e/ directory and shell execution via npx (documented in SKILL.md).
Sanitization: Not instructed; the skill relies on the agent to treat the input as a simple string.

create-regression-test