Merge origin/main into fix/pre-wordpress-request-dispatch

Author: chubes4

.github/workflows/agent-task-contracts.yml

@@ -96,6 +96,7 @@ jobs:
       - run: npm run test:native-agent-task-playground-e2e
       - run: npm run test:native-agent-task-interruption
       - run: npm run test:trusted-apply-artifact-channel
+      - run: npm run test:runtime-command-artifact-bounds
       - run: npm run test:redaction
       - run: npm run test:production-boundary-enforcement
       - run: npm run test:runtime-tool-policy

CHANGELOG.md

@@ -2,6 +2,208 @@
 
 All notable changes to this project will be documented in this file. Entries are generated by homeboy from git commits.
 
+## [0.15.0] - 2026-07-25
+
+### Added
+- exercise AJAX and XML-RPC adversarial surfaces
+- add external MySQL runtime provider
+- complete supported adversarial campaigns
+- add WordPress adversarial Playground adapter
+- explore adaptive browser state graphs
+- run adversarial campaigns from recipes
+
+### Changed
+- cover generated runtime secret evidence
+
+### Fixed
+- require explicit plugin composer preparation
+- stabilize adaptive replay across volatile ids
+- enforce external database security boundaries
+- exclude dependency trees from readonly mounts
+- enforce adversarial HTTP action contracts
+- honor routed preview hosts in adaptive exploration
+- keep failed adaptive actions off frontier
+- generate unique action corpus selectors
+- fail closed on unapplied fault schedules
+
+## [0.14.0] - 2026-07-24
+
+### Added
+- add adversarial fuzz runtime contracts
+
+### Changed
+- Handle cross-realm PHP-WASM failures
+- Narrow Playground worker terminalization
+- Terminate poisoned Playground workers
+- Exercise queued D1 artifact imports
+- Recover completed publication promotion
+- Complete provisioned site publication lifecycle
+- Accept external provisioning artifacts in local gate
+- Add public provisioning local gate
+- support native WP-CLI stress runner
+- Add authenticated provisioning artifact staging
+
+### Fixed
+- Fix readonly mount symlink cycles
+- Fix PHPUnit WASM runtime rejection handling
+- preserve deferred WordPress hook lifecycle
+- defer WP-CLI wrapper cleanup to runtime owner
+- isolate concurrent WP-CLI wrapper scripts
+
+## [0.13.2] - 2026-07-23
+
+### Changed
+- Add one-time administrator claims
+- Add authenticated Cloudflare provisioning API core
+- Add durable Cloudflare operation lifecycle
+- Preserve lossless D1 rollback state
+- Preserve Durable Object class during D1 cutover
+- Stream large staged file materialization
+- Preserve glob strings in D1 config
+- Make emitted D1 configs portable
+- Provision Cloudflare D1 coordinator deterministically
+- Scope Cloudflare state by site identity
+- Import static artifacts into Cloudflare runtime
+- Fence Cloudflare coordinator cutovers
+- Adopt exact Cloudflare coordinator state
+- Reuse canonical publication snapshots
+- Bound Cloudflare publication scheduler CPU
+- Bound Cloudflare mutation memory
+- Drain Cloudflare publication in bounded jobs
+- Require explicit Cloudflare publication changes
+- Reconcile dependency overlay to consumer autoload layout
+
+### Fixed
+- force fresh dist for release packages
+- honor MySQL contracts in managed PHPUnit
+- avoid inactive plugin dev autoloaders
+- contain readonly mount symlinks
+
+## [0.13.1] - 2026-07-22
+
+### Changed
+- Publish affected Cloudflare routes automatically (by Chris Huber)
+- Make Cloudflare revision coordination configurable (by Chris Huber)
+- Serve published WordPress routes without coordination (by Chris Huber)
+
+### Fixed
+- Fix release artifact installability (by Chris Huber)
+
+## [0.13.0] - 2026-07-21
+
+### Changed
+- Persist mutable WordPress content in R2 (by Chris Huber)
+- Exercise canonical Cloudflare boot probe (by Chris Huber)
+- Bound retained recipe command evidence ([#1947](https://github.com/Automattic/wp-codebox/pull/1947)) (by Chris Huber)
+- Navigate multi-actor browser sessions ([#1948](https://github.com/Automattic/wp-codebox/pull/1948)) (by Chris Huber)
+- Load Cloudflare runtime dependencies from R2 (by Chris Huber)
+- Serve WordPress assets from R2 ranges (by Chris Huber)
+- Persist Cloudflare WordPress media in R2 (by Chris Huber)
+- Add PHPUnit config to readonly cache fixture ([#1942](https://github.com/Automattic/wp-codebox/pull/1942)) (by Chris Huber)
+- Allow managed PHPUnit discovery without config ([#1944](https://github.com/Automattic/wp-codebox/pull/1944)) (by Chris Huber)
+- Declare PHPUnit config in readonly fixture (by Chris Huber)
+- Serve cached Cloudflare pages before runtime boot (by Chris Huber)
+- Cache public pages by canonical revision (by Chris Huber)
+- Fail closed on PHPUnit discovery errors ([#1940](https://github.com/Automattic/wp-codebox/pull/1940)) (by Chris Huber)
+- Add canonical Cloudflare state restore (by Chris Huber)
+- Capture malformed PHPUnit bootstrap payloads ([#1933](https://github.com/Automattic/wp-codebox/pull/1933)) (by Chris Huber)
+- Harden Playground cache retention safety ([#1930](https://github.com/Automattic/wp-codebox/pull/1930)) (by Chris Huber)
+- Retain non-fatal PHPUnit bootstrap exits ([#1932](https://github.com/Automattic/wp-codebox/pull/1932)) (by Chris Huber)
+- Align Playground runtime bootstrap dependencies ([#1931](https://github.com/Automattic/wp-codebox/pull/1931)) (by Chris Huber)
+- Repair Cloudflare WordPress corpus dev rig (by Chris Huber)
+- Execute bounded plans through one recipe runtime ([#1928](https://github.com/Automattic/wp-codebox/pull/1928)) (by Chris Huber)
+- Run bounded PHPUnit plans in one disposable runtime ([#1922](https://github.com/Automattic/wp-codebox/pull/1922)) (by Chris Huber)
+- Preserve binary staged file bytes ([#1921](https://github.com/Automattic/wp-codebox/pull/1921)) (by Chris Huber)
+- Retain PHPUnit bootstrap failure diagnostics ([#1918](https://github.com/Automattic/wp-codebox/pull/1918)) (by Chris Huber)
+- Preserve Playground PHPUnit crash diagnostics ([#1917](https://github.com/Automattic/wp-codebox/pull/1917)) (by Chris Huber)
+- Add canonical Cloudflare runtime homepage (by Chris Huber)
+- Restore hook snapshot iteration (by Chris Huber)
+- Preserve generic hook removal behavior (by Chris Huber)
+- Remove Cloudflare investigation diagnostics (by Chris Huber)
+- Revert "Add bounded MDI init diagnostics" (by Chris Huber)
+- Revert "Probe complete canonical WordPress state" (by Chris Huber)
+- Revert "Probe canonical lifecycle boundaries" (by Chris Huber)
+- Revert "Probe MDI auto-draft persistence" (by Chris Huber)
+- Revert "Log canonical bootstrap stages" (by Chris Huber)
+- Revert "Persist canonical bootstrap stage evidence" (by Chris Huber)
+- Add coordinated multi-actor browser scenarios ([#1913](https://github.com/Automattic/wp-codebox/pull/1913)) (by Chris Huber)
+- Revert "Persist mutation stage evidence" (by Chris Huber)
+- Pin bounded MDI option persistence (by Chris Huber)
+- Probe shared MDI temp paths (by Chris Huber)
+- Isolate MDI option mutation tracking (by Chris Huber)
+- Isolate MDI dirty option writes (by Chris Huber)
+- Isolate MDI options fetch (by Chris Huber)
+- Isolate MDI options index maintenance (by Chris Huber)
+- Isolate MDI table flush probes (by Chris Huber)
+- Report MDI dirty flush inventory (by Chris Huber)
+- Split MDI write lifecycle probes (by Chris Huber)
+- Support external MariaDB PHPUnit runtimes ([#1912](https://github.com/Automattic/wp-codebox/pull/1912)) (by Chris Huber)
+- Allow mounts to skip artifact capture ([#1910](https://github.com/Automattic/wp-codebox/pull/1910)) (by Chris Huber)
+- Add MDI persistence isolation probes (by Chris Huber)
+- Recycle PHP at canonical request boundaries (by Chris Huber)
+- Instrument editor footer scripts (by Chris Huber)
+- Instrument block editor output phases (by Chris Huber)
+- Split canonical bootstrap setup phases (by Chris Huber)
+- Seed custom CSS lookup cache canonically (by Chris Huber)
+- Seed WordPress custom CSS absence cache (by Chris Huber)
+- Anchor custom CSS fast path uniquely (by Chris Huber)
+- Fast-path custom CSS theme JSON output (by Chris Huber)
+- Compare editor custom and theme styles (by Chris Huber)
+- Split global stylesheet memory phases (by Chris Huber)
+- Instrument core editor settings phases (by Chris Huber)
+- Instrument block editor settings memory (by Chris Huber)
+- Probe block editor global styles lookup (by Chris Huber)
+- Instrument block editor REST preload memory (by Chris Huber)
+- Split editor post-insert memory phases (by Chris Huber)
+- Tighten editor insertion memory marker (by Chris Huber)
+- Instrument Cloudflare editor memory phases (by Chris Huber)
+- Package constrained MDI runtime dependencies (by Chris Huber)
+- Bundle upstream auto-draft persistence fix (by Chris Huber)
+- Add bounded Cloudflare editor probes (by Chris Huber)
+- Honor coordinator lease retry timing (by Chris Huber)
+- Prevent stale runner publication branches ([#1895](https://github.com/Automattic/wp-codebox/pull/1895)) (by Chris Huber)
+- Separate workflow helpers from release assets ([#1892](https://github.com/Automattic/wp-codebox/pull/1892)) (by Chris Huber)
+- Preserve verification failure evidence ([#1890](https://github.com/Automattic/wp-codebox/pull/1890)) (by Chris Huber)
+- Document canonical lifecycle probes (by Chris Huber)
+- Add declared verification command artifacts ([#1888](https://github.com/Automattic/wp-codebox/pull/1888)) (by Chris Huber)
+- Persist canonical MDI change sets (by Chris Huber)
+- Add authenticated canonical state reset (by Chris Huber)
+- Parallelize canonical R2 persistence (by Chris Huber)
+- Keep rewrite cache out of canonical state (by Chris Huber)
+- Probe rewrite rule persistence (by Chris Huber)
+- Add direct wp-loaded probe aliases (by Chris Huber)
+- Add short canonical lifecycle probes (by Chris Huber)
+- Bisect canonical wp-loaded callbacks (by Chris Huber)
+- Disable browser-request cron scheduling (by Chris Huber)
+- Patch disabled cron before canonical init (by Chris Huber)
+- Remove disabled cron at init time (by Chris Huber)
+- Skip disabled cron registration in Cloudflare (by Chris Huber)
+- Isolate canonical cron scheduling (by Chris Huber)
+- Remove diagnostic hooks through WordPress APIs (by Chris Huber)
+- Probe remaining widget registrations (by Chris Huber)
+- Compare MDI widget option visibility (by Chris Huber)
+- Bisect MDI widget settings registration (by Chris Huber)
+- Split widget initialization diagnostics (by Chris Huber)
+- Extract the runtime corpus inside PHP (by Chris Huber)
+- Buffer the bounded R2 runtime artifact (by Chris Huber)
+- Provision the WordPress runtime corpus in R2 (by Chris Huber)
+- Stream WordPress browser assets outside PHP (by Chris Huber)
+- Bound the Cloudflare WordPress runtime (by Chris Huber)
+- Keep PHP outside the state coordinator (by Chris Huber)
+- Prebuild canonical Cloudflare WordPress state (by Chris Huber)
+- Persist functional WordPress state in Cloudflare (by Chris Huber)
+
+### Fixed
+- Fix staged binary assets in browser previews ([#1939](https://github.com/Automattic/wp-codebox/pull/1939)) (by Chris Huber)
+- Fix Cloudflare site styles and warm reads (by Chris Huber)
+- Fix Playground cache leases on macOS ([#1936](https://github.com/Automattic/wp-codebox/pull/1936)) (by Chris Huber)
+- bound Playground custom archive cache ([#1929](https://github.com/Automattic/wp-codebox/pull/1929)) (by Chris Huber)
+- support private PHPUnit TestSuite constructors ([#1927](https://github.com/Automattic/wp-codebox/pull/1927)) (by Chris Huber)
+- expose CLI streams in Playground PHP runtimes ([#1905](https://github.com/Automattic/wp-codebox/pull/1905)) (by Chris Huber)
+- decode complete visual compare images ([#1907](https://github.com/Automattic/wp-codebox/pull/1907)) (by Chris Huber)
+- wait for readonly mount staging before cleanup ([#1897](https://github.com/Automattic/wp-codebox/pull/1897)) (by Chris Huber)
+- provision multisite PHPUnit runtimes ([#1900](https://github.com/Automattic/wp-codebox/pull/1900)) (by Chris Huber)
+
 ## [0.12.29] - 2026-07-18
 
 ### Changed

README.md

@@ -1,5 +1,9 @@
 # WP Codebox
 
+Adversarial campaign, transport fault, service disruption, browser oracle, and
+sealed replay contracts are documented in
+[`docs/adversarial-runtime.md`](docs/adversarial-runtime.md).
+
 **WP Codebox unlocks secure WordPress code execution from anywhere.** Run agents, accept untrusted patches, evaluate plugins, reproduce bugs, or experiment freely - every sandbox is a disposable contained WordPress runtime that can't touch its caller. Your host can be a CLI, CI job, mobile app, Node service, WordPress plugin, or anything else that can shell out or hit an API.
 
 WordPress has historically lacked a clean scratch space for code execution. Modern dev workflows assume one - Node has `npm install` per project, Python has venvs, containers have ephemeral filesystems. WP Codebox provides that primitive as a usable runtime contract: real WordPress, fully ephemeral, no host filesystem access except via declared mounts. Any product - WordPress or not - can offer code execution against a real WordPress instance without risking the caller.
@@ -721,7 +725,7 @@ Supported runtime commands today:
 - `wordpress.server-page-load`: load `surface=admin|frontend` plus `path=<path>` or `url=<path-or-url>` through the live preview HTTP server without a browser. Its performance observation is marked `source=server-http` and `kind=server-page-load`.
 - `wordpress.browser-page-load`: load `surface=admin|frontend` plus `path=<path>` or `url=<path-or-url>` in Playwright by wrapping `wordpress.browser-probe`.
 - `wordpress.browser-probe`: boot the live preview, visit `url=<path-or-url>` with Playwright, and capture generic browser replay/audit evidence under `files/browser/`.
-- `wordpress.browser-actions`: boot the live preview, drive it with an ordered interaction script (`steps-json`), assert browser behavior, and capture replay/audit evidence under `files/browser/`.
+- `wordpress.browser-actions`: boot the live preview, drive it with an ordered interaction script (`steps-json`) or bounded adaptive exploration, optionally classify accessibility/keyboard/focus failures, and capture replay/audit evidence under `files/browser/`.
 
 `wordpress.run-php` loads `/wordpress/wp-load.php` by default. Use `--arg bootstrap=none` for raw PHP.
 
@@ -1305,7 +1309,7 @@ Callers can gate success on post-agent verification through `verify_steps`: an a
 
 Each component contract declares `slug`, `path` or `source`, optional explicit `pluginFile`, optional `activate`, optional `loadAs`, and optional `readiness_probe` metadata.
 
-The CLI binary can come from ability input, the `wp_codebox_bin` option, or the `wp_codebox_bin` filter. Source-checkout snapshots that do not include generated `dist/` files should point at `bin/wp-codebox-source.mjs`; it builds WP Codebox when needed before delegating to the compiled CLI.
+The CLI binary can come from ability input, the `wp_codebox_bin` option, or the `wp_codebox_bin` filter. Source-checkout snapshots should point at `bin/wp-codebox-source.mjs`; it runs the incremental source build before every invocation so ignored `dist/` output cannot predate the checked-out core, Playground, or CLI source, then delegates to the compiled CLI.
 
 Caller-owned runtime components can provide workspace, file, GitHub, or other tools to the sandboxed agent. WP Codebox owns the parent-site ability surface, sandbox lifecycle, and artifact capture boundary.
 

bin/wp-codebox-source.mjs

@@ -35,18 +35,21 @@ function run(command, args, options = {}) {
   }
 
   if (result.status !== 0) {
+    if (options.failureContext) {
+      console.error(`WP Codebox source entrypoint failed to ${options.failureContext} (exit ${result.status ?? 1}).`)
+    }
     process.exit(result.status ?? 1)
   }
 }
 
-if (!existsSync(distEntrypoint)) {
+const distIsAbsent = !existsSync(distEntrypoint)
+if (distIsAbsent) {
   console.error("WP Codebox CLI dist entrypoint is absent; bootstrapping the source checkout before running the CLI.")
+}
 
-  if (!existsSync(nodeModules)) {
-    run("npm", [existsSync(packageLock) ? "ci" : "install"])
-  }
-
-  run("npm", ["run", "build"])
+if (!existsSync(nodeModules)) {
+  run("npm", [existsSync(packageLock) ? "ci" : "install"], { failureContext: "install source checkout dependencies" })
 }
 
+run("npm", ["run", "build"], { failureContext: "build the source checkout" })
 run(process.execPath, [distEntrypoint, ...process.argv.slice(2)], { cwd: callerCwd, delegate: true })

docs/README.md

@@ -41,6 +41,9 @@ unless this index says otherwise.
 - [Generic runtime primitives](./generic-runtime-primitives.md) documents the
   caller-neutral artifact storage, trusted browser origin, materialization, and
   target-context envelopes shared by runtime integrations.
+- [Browser environment matrices](./browser-environment-matrices.md) documents
+  deterministic hostile-environment expansion, capability fidelity, bounds,
+  isolated cell artifacts, and replay evidence.
 - [Runtime profile contract](./runtime-profile-contract.md) documents
   `wp-codebox/runtime-profile/v1`, the Codebox-owned profile request/result lane
   for capabilities, components, readiness, diagnostics, and provenance.