name: initialize-testing description: Set up an adopting project's Markdown test catalog, lifecycle functions, runtime manifest, and evidence review configuration for SnagBadger testing.#

Read the testing contract from the SnagBadger runner root. Use the project's existing test directory/configuration when present. Create only the missing configuration and useful example scenarios; preserve authored tests.

Find the runner installation and verify its pinned toolchain/dependencies/build. Discover the user's available agent-delegation and browser MCP capabilities. Keep the runtime portable: configure the host's tools, not a mandatory T3/browser product. Browser execution requires a fresh context per attempt and scoped action/cleanup support. Record missing capabilities explicitly; required video cannot be replaced by a screenshot without changing the project's evidence settings.

Set up a directory catalog of .snag.md tests. Recommend explicit section-local IDs, executor type, prerequisites, checkbox steps, and expected outcomes; free-form Markdown is valid. Add section.md only for shared metadata, guidance, hooks, or destination settings. Demonstrate tag selection when relevant.

Register project functions in an explicitly selected module. A per-test authenticateUser function uses context.browser.perform to request actions in the already-created isolated context, or performs API/CLI setup directly. Run/section hooks prepare shared resources; do not use a once-per-section browser login as per-test authentication. Read inputs/credentials from the project environment rather than frontmatter or exported configuration.

Configure reportPolicy through configureTesting helpers or a module export. Review complete report text/source context and attachment bytes using the existing sanitization boundary. Only include screenshots/video the policy can review; omissions remain visible. Exact-string secret removal is a helper, not binary sanitization or automatic privacy detection.

Run inspect and demonstrate the smallest relevant scenario through run-test-plan. Report the actual catalog/runtime configuration and any blocked capabilities. Installing plugins, changing global MCP configuration, deployment, and live tracker submission are outside project setup unless separately requested.