The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the TestSprite CLI via npm exec with pinned versions (e.g., @testsprite/testsprite-cli@0.4.0) to perform network-bound cloud testing and local file operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes external test plans provided via <PLAN_FILE>. It employs a mandatory evidence chain for safety: instructions require reading repository authority, following non-negotiable safety rules, and performing automated sanitization using a custom Python auditor (audit_frontend_plan.py) and vendor linting tools to detect malicious patterns or bypasses before execution.
[CREDENTIALS_UNSAFE]: Handles the TESTSPRITE_API_KEY but implements strict mitigations, including mandatory umask 077 for temporary storage, the use of env -u to prevent accidental persistence, and an auditing script that blocks literal credentials (tokens, JWTs, keys) from being included in test plans.
[DYNAMIC_EXECUTION]: Facilitates the retrieval and modification of Playwright Python scripts (test code put). Security is maintained through strict version verification and instructions for manual code review by the agent to prevent the introduction of unverified or malicious logic.

run-testsprite-frontend