The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from SAP Analytics Cloud dashboards during the discovery phase.
Ingestion points: Browser snapshots, table data, and widget metadata from SAC stories (SKILL.md, references/agent-browser-discovery.md).
Boundary markers: The skill explicitly enforces a "discovery proposes, humans approve" rule to prevent autonomous execution of discovered patterns (SKILL.md).
Capability inventory: File system access to write discovery artifacts and subprocess execution for browser drivers (references/architecture.md).
Sanitization: Requires manual inspection of all generated profile and scenario drafts before they are used in CI pipelines (references/dashboard-profiles-and-scenarios.md).
[DYNAMIC_EXECUTION]: Generates executable test configurations and scenarios from templates.
Template usage: Assembles dashboard-profile.yaml and scenario-read-only-smoke.yaml based on metadata gathered during the discovery phase (SKILL.md, templates/).
Execution context: Generated artifacts are intended for execution via the Playwright framework (references/playwright-execution.md).
[COMMAND_EXECUTION]: Utilizes shell commands for environment discovery and tool orchestration.
Environment checks: Includes PowerShell and shell commands to verify the presence of Node.js, npm, and browser executables (references/tool-availability-and-deployment.md).
Tool invocation: Uses npx to launch the Chrome DevTools MCP server (references/chrome-devtools-mcp.md).
[EXTERNAL_DOWNLOADS]: References several external browser automation tools and Node.js packages.
Dependency list: Recommends playwright, agent-browser, and chrome-devtools-mcp (SKILL.md).
Safety controls: Advises using pinned versions, internal registries, and disabling usage statistics/telemetry for these tools (references/tool-availability-and-deployment.md).

sap-sac-test-automation