The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The onboarding process involves reading project context from external files, which could serve as a vector for indirect prompt injection if those files contain malicious instructions.\n
Ingestion points: The skill reads .agents/qa-project-context.md, CLAUDE.md, and AGENTS.md during the orientation phase in SKILL.md.\n
Boundary markers: The instructions do not specify using delimiters or explicit warnings to ignore potentially malicious content within these files.\n
Capability inventory: The skill provides instructions for executing tests via shell commands and writing documentation to the repository.\n
Sanitization: No content filtering or sanitization steps are defined for the data read from external project files.\n- [COMMAND_EXECUTION]: The skill provides comprehensive templates for executing Playwright test commands, which is the standard method for running tests in this framework.\n
Evidence: Includes commands for running tests (npx playwright test), debugging (--debug), and viewing reports in references/framework-walkthrough.md.\n
Context: These commands are standard for the Playwright framework and align with the skill's purpose of test automation and onboarding.\n- [EXTERNAL_DOWNLOADS]: The skill references the Playwright framework and associated CLI tools for automation.\n
Evidence: Mentions @playwright/cli and playwright-cli in the context of agent-driven automation in SKILL.md.\n
Context: Playwright is a well-known tool from Microsoft. The use of non-standard package names in the documentation is noted but does not involve forced execution or unverified installation.

qa-project-bootstrap