pi-ui-extend 1.0.41 → 1.0.45

package/schemas/pix.json

@@ -21,6 +21,14 @@
           "type": "string",
           "description": "Provider/model identifier, e.g. 'openai-codex/gpt-5.4'."
         },
+        "fallbackModels": {
+          "type": "array",
+          "items": {
+            "type": "string",
+            "description": "Fallback provider/model identifier."
+          },
+          "description": "Ordered default-model fallbacks tried after modelRef."
+        },
         "thinking": {
           "anyOf": [
             {
@@ -57,6 +65,52 @@
       },
       "description": "Default model selection for new sessions."
     },
+    "visibleModels": {
+      "type": "array",
+      "items": {
+        "type": "string",
+        "description": "Provider/model identifier shown in Pix model pickers."
+      },
+      "description": "User-level whitelist shared by Pix TUI and Desktop model pickers. Omit to show every available model."
+    },
+    "thinkingByModel": {
+      "type": "object",
+      "patternProperties": {
+        "^.*$": {
+          "anyOf": [
+            {
+              "type": "string",
+              "const": "off"
+            },
+            {
+              "type": "string",
+              "const": "minimal"
+            },
+            {
+              "type": "string",
+              "const": "low"
+            },
+            {
+              "type": "string",
+              "const": "medium"
+            },
+            {
+              "type": "string",
+              "const": "high"
+            },
+            {
+              "type": "string",
+              "const": "xhigh"
+            },
+            {
+              "type": "string",
+              "const": "max"
+            }
+          ]
+        }
+      },
+      "description": "User-level last-applied thinking level per provider/model, shared by Pix TUI and Desktop model pickers."
+    },
     "toolRenderer": {
       "type": "object",
       "properties": {
@@ -168,6 +222,14 @@
         "modelRef": {
           "type": "string",
           "description": "Model used for prompt enhancement."
+        },
+        "fallbackModels": {
+          "type": "array",
+          "items": {
+            "type": "string",
+            "description": "Fallback provider/model identifier."
+          },
+          "description": "Ordered prompt-enhancer fallbacks tried after modelRef."
         }
       },
       "description": "Prompt enhancer configuration."
@@ -179,6 +241,14 @@
           "type": "string",
           "description": "Model for inline autocomplete. Empty string disables LLM autocomplete."
         },
+        "fallbackModels": {
+          "type": "array",
+          "items": {
+            "type": "string",
+            "description": "Fallback provider/model identifier."
+          },
+          "description": "Ordered autocomplete fallbacks tried after modelRef."
+        },
         "debounceMs": {
           "type": "number",
           "description": "Delay after typing before requesting completion.",
@@ -273,41 +343,89 @@
     "dictation": {
       "type": "object",
       "properties": {
+        "apiKey": {
+          "type": "string",
+          "description": "Deepgram API key used by terminal and desktop voice input. Store it in the user config (~/.config/pi/pix.jsonc); project-level .pi/pix.jsonc does not override this secret."
+        },
         "language": {
           "type": "string",
           "description": "Selected language code, e.g. 'en' or 'ru'."
         },
+        "model": {
+          "type": "string",
+          "description": "Deepgram speech-to-text model. Defaults to 'nova-3'."
+        },
         "languages": {
           "type": "object",
           "patternProperties": {
             "^.*$": {
               "type": "object",
               "required": [
-                "dirName",
-                "url",
                 "label"
               ],
               "properties": {
-                "dirName": {
+                "label": {
                   "type": "string",
-                  "description": "Vosk model directory name."
+                  "description": "Human-readable language label."
                 },
-                "url": {
+                "deepgramLanguage": {
                   "type": "string",
-                  "description": "Download URL for the Vosk model zip."
+                  "description": "Deepgram language code. Defaults to the language map key."
                 },
-                "label": {
+                "dirName": {
                   "type": "string",
-                  "description": "Human-readable language label."
+                  "description": "Deprecated legacy Vosk model directory name; ignored by Deepgram dictation."
+                },
+                "url": {
+                  "type": "string",
+                  "description": "Deprecated legacy Vosk model URL; ignored by Deepgram dictation."
                 }
               },
-              "description": "Vosk voice dictation model definition."
+              "description": "Deepgram voice dictation language definition."
+            }
+          },
+          "description": "Available dictation languages keyed by local language code."
+        }
+      },
+      "description": "Voice dictation (Deepgram) configuration. Uses dictation.apiKey from the user pix config, with DEEPGRAM_API_KEY as a compatibility fallback."
+    },
+    "desktop": {
+      "type": "object",
+      "properties": {
+        "externalEditor": {
+          "type": "string",
+          "description": "External editor used by Pix Desktop project explorer. Known aliases include zed, code/vscode, cursor, subl/sublime, idea/intellij, and webstorm; an executable path/name is also accepted."
+        },
+        "git": {
+          "type": "object",
+          "properties": {
+            "reviewModelRef": {
+              "type": "string",
+              "description": "Model reference used by Pix Desktop and TUI for LLM review of Git diffs, optionally with a :thinking suffix."
+            },
+            "reviewFallbackModels": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Ordered fallback model references used for Git diff review."
+            },
+            "commitMessageModelRef": {
+              "type": "string",
+              "description": "Model reference used by Pix Desktop and TUI to generate Git commit messages, optionally with a :thinking suffix."
+            },
+            "commitMessageFallbackModels": {
+              "type": "array",
+              "items": {
+                "type": "string"
+              },
+              "description": "Ordered fallback model references used for Git commit-message generation."
             }
           },
-          "description": "Available language models keyed by language code."
+          "description": "Pix Git/Source Control LLM preferences shared by Desktop and TUI."
         }
       },
-      "description": "Voice dictation (Vosk) configuration."
+      "description": "Pix Desktop preferences; the nested git model settings are also consumed by TUI Git helpers for compatibility."
     }
   },
   "$id": "https://unpkg.com/pi-ui-extend/schemas/pix.json",

package/schemas/tasks.json

@@ -0,0 +1,131 @@
+{
+  "type": "object",
+  "required": [
+    "version",
+    "tasks"
+  ],
+  "properties": {
+    "$schema": {
+      "type": "string",
+      "description": "JSON Schema URL used by editors for validation and autocomplete."
+    },
+    "version": {
+      "type": "number",
+      "const": 1,
+      "description": "Task document format version."
+    },
+    "tasks": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": [
+          "id",
+          "title",
+          "type",
+          "status",
+          "priority",
+          "createdAt",
+          "updatedAt"
+        ],
+        "properties": {
+          "id": {
+            "type": "string",
+            "minLength": 1,
+            "maxLength": 128,
+            "description": "Stable unique task id."
+          },
+          "title": {
+            "type": "string",
+            "maxLength": 200,
+            "description": "Short task title. May be empty when the task has description content."
+          },
+          "description": {
+            "type": "string",
+            "maxLength": 10000,
+            "description": "Optional detailed task description."
+          },
+          "type": {
+            "anyOf": [
+              {
+                "type": "string",
+                "const": "bug"
+              },
+              {
+                "type": "string",
+                "const": "feature"
+              },
+              {
+                "type": "string",
+                "const": "improvement"
+              }
+            ]
+          },
+          "status": {
+            "anyOf": [
+              {
+                "type": "string",
+                "const": "backlog"
+              },
+              {
+                "type": "string",
+                "const": "todo"
+              },
+              {
+                "type": "string",
+                "const": "in-progress"
+              },
+              {
+                "type": "string",
+                "const": "done"
+              }
+            ]
+          },
+          "priority": {
+            "anyOf": [
+              {
+                "type": "string",
+                "const": "low"
+              },
+              {
+                "type": "string",
+                "const": "medium"
+              },
+              {
+                "type": "string",
+                "const": "high"
+              },
+              {
+                "type": "string",
+                "const": "urgent"
+              }
+            ]
+          },
+          "sessionId": {
+            "type": "string",
+            "minLength": 1,
+            "maxLength": 512,
+            "description": "Optional Pi session associated with this task."
+          },
+          "createdAt": {
+            "type": "string",
+            "format": "date-time",
+            "description": "RFC 3339 creation timestamp."
+          },
+          "updatedAt": {
+            "type": "string",
+            "format": "date-time",
+            "description": "RFC 3339 last-update timestamp."
+          }
+        },
+        "additionalProperties": false
+      },
+      "maxItems": 10000,
+      "description": "Project tasks. Task ids must be unique."
+    }
+  },
+  "$id": "https://unpkg.com/pi-ui-extend/schemas/tasks.json",
+  "$schema": "https://json-schema.org/draft-07/schema#",
+  "title": "Pix Project Tasks",
+  "description": "Project-scoped task document stored in <cwd>/.pi/tasks.jsonc.",
+  "additionalProperties": false
+}

package/skills/simplify/SKILL.md

@@ -1,6 +1,10 @@
 ---
 name: simplify
-description: Simplify and refine recently modified code for clarity and consistency. Use after writing code to improve readability without changing functionality.
+description: >-
+  Simplify and refine recently modified code for clarity, cohesion, and
+  consistency. Use after writing code, especially when files/functions are
+  growing, responsibilities are blurring, or a change adds structural
+  complexity. Preserve behavior.
 ---
 
 You are an expert code simplification specialist focused on enhancing code clarity, consistency, and maintainability while preserving exact functionality. Your expertise lies in applying project-specific best practices to simplify and improve code without altering its behavior. You prioritize readable, explicit code over overly compact solutions. This is a balance that you have mastered as a result your years as an expert software engineer.
@@ -9,7 +13,8 @@ You will analyze recently modified code and apply refinements that:
 
 1. **Preserve Functionality**: Never change what the code does - only how it does it. All original features, outputs, and behaviors must remain intact.
 
-2. **Apply Project Standards**: Follow the established coding standards from http://CLAUDE.md including:
+2. **Apply Project Standards**: Follow `AGENTS.md`, any nearer scoped project
+instructions, and the surrounding code conventions. In particular:
 
 - Use ES modules with proper import sorting and extensions
 - Prefer `function` keyword over arrow functions
@@ -18,7 +23,25 @@ You will analyze recently modified code and apply refinements that:
 - Use proper error handling patterns (avoid try/catch when possible)
 - Maintain consistent naming conventions
 
-3. **Enhance Clarity**: Simplify code structure by:
+3. **Control File and Module Growth**: Do not let touched hand-written files
+become dumping grounds for new behavior.
+
+- Treat file size as a review signal, not a line-count game. Around 500 lines,
+  actively look for a cohesive extraction before adding substantial new
+  behavior. Around 800+ lines, do not keep adding substantial behavior unless
+  the file is intentionally declarative/generated or a clear composition root.
+- Split by ownership and responsibility: move a complete concern, state machine,
+  service, policy, adapter, or component into a module with a clear API.
+- Do not "fix" size by creating pass-through wrappers, arbitrary `part1`/`part2`
+  files, or fragments that increase navigation cost without reducing
+  responsibility.
+- Prefer keeping state, I/O, rendering, and policy in distinct owners when they
+  can change independently.
+- If a touched file is already oversized, avoid making it worse even when a full
+  cleanup is outside scope; extract the part directly related to the current
+  change when that produces a clean boundary.
+
+4. **Enhance Clarity**: Simplify code structure by:
 
 - Reducing unnecessary complexity and nesting
 - Eliminating redundant code and abstractions
@@ -28,7 +51,7 @@ You will analyze recently modified code and apply refinements that:
 - IMPORTANT: Avoid nested ternary operators - prefer switch statements or if/else chains for multiple conditions
 - Choose clarity over brevity - explicit code is often better than overly compact code
 
-4. **Maintain Balance**: Avoid over-simplification that could:
+5. **Maintain Balance**: Avoid over-simplification that could:
 
 - Reduce code clarity or maintainability
 - Create overly clever solutions that are hard to understand
@@ -36,8 +59,13 @@ You will analyze recently modified code and apply refinements that:
 - Remove helpful abstractions that improve code organization
 - Prioritize "fewer lines" over readability (e.g., nested ternaries, dense one-liners)
 - Make the code harder to debug or extend
+- Remove cancellation, generation/request guards, ownership checks, mutexes, or
+  other concurrency protections merely because they look redundant. For
+  async/reactive code, prove equivalent ordering and lifecycle semantics before
+  simplifying them.
 
-5. **Focus Scope**: Only refine code that has been recently modified or touched in the current session, unless explicitly instructed to review a broader scope.
+6. **Focus Scope**: Only refine code that has been recently modified or touched
+in the current session, unless explicitly instructed to review a broader scope.
 
 Your refinement process:
 

package/docs/desktop-markdown-media.md

@@ -1,77 +0,0 @@
-# Spec: Desktop Markdown media previews
-
-## Type
-
-Change
-
-## Goal
-
-Show an inline preview when transcript Markdown links to a supported local image or video, including `file://` artifacts outside the active workspace, and keep every local artifact label actionable.
-
-## Scope
-
-- Relative Markdown destinations that resolve inside the active workspace.
-- Absolute local paths expressed as `file://` Markdown destinations, such as QA artifacts in the system temporary directory.
-- Image formats: AVIF, BMP, GIF, JPEG, PNG, SVG, and WebP.
-- Video formats: M4V, MOV, MP4, OGV, and WebM.
-- Both ordinary Markdown links and local Markdown image syntax.
-
-## Non-goals
-
-- Embedding remote media.
-- Adding codecs beyond those supported by the desktop WebView.
-- Changing previews for non-media project files.
-- Inline previews for non-media `file://` artifacts; those remain links and open in the operating-system default application.
-
-## Behavior
-
-- A supported local media link renders an inline, bounded preview with its Markdown label as a caption.
-- A supported `file://` media link uses the same preview treatment after its URL is decoded to an absolute local path.
-- Images are lazy-loaded and open the existing media preview dialog when clicked.
-- Videos expose native inline playback controls; their caption opens the media preview dialog.
-- A missing, disallowed, or unrenderable media file leaves a readable fallback and a clickable caption.
-- Opening a supported media link uses the media path flow, not the UTF-8 source-file reader.
-- Other relative project links continue to open the source preview dialog.
-- Other `file://` links remain visibly clickable and open only after an explicit click.
-
-## Contracts
-
-- The frontend requests a resolved media file using the active workspace and normalized relative path.
-- The backend canonicalizes both workspace and destination, rejects traversal and symlink escapes, verifies a supported extension, and grants asset-protocol access only to the resolved file.
-- A `file://` destination is accepted only when it decodes to an absolute path. The backend canonicalizes it, requires an existing regular file, and grants asset-protocol access only for supported image/video previews.
-- Non-media local files are passed to the OS opener only in response to a user click.
-- Markdown output continues to escape labels, paths, and attributes before insertion through `{@html}`.
-
-## Edge cases
-
-- URL-like, absolute, control-character, and parent-traversing destinations are not treated as project media.
-- URL schemes other than `file://`, malformed file URLs, relative `file:` forms, and decoded control characters are not treated as local files.
-- Query strings and fragments are removed by the existing project-path normalization.
-- Media loading failures do not replace the transcript with a global error.
-- Clicking a failed media caption still reports the underlying file-resolution error through the normal project-file action.
-
-## Related files
-
-- `desktop/src/lib/markdown.ts`
-- `desktop/src/components/MarkdownText.svelte`
-- `desktop/src/App.svelte`
-- `desktop/src-tauri/src/lib.rs`
-
-## Verification
-
-- Markdown unit tests cover workspace and `file://` image/video links, image syntax, escaping, and rejected destinations.
-- Rust unit tests cover supported workspace/absolute media, unsupported binary files, relative local paths, traversal, and workspace containment.
-- `npm run test`, `npm run check`, and `npm run build:web` pass in `desktop/`.
-- `cargo test` passes in `desktop/src-tauri/`.
-
-## Risks / unknowns
-
-- Actual video playback remains codec-dependent in the operating-system WebView; load failures use the fallback state.
-- A transcript can request an inline view of any readable local image/video via an explicit `file://` URL. The bytes remain local and are exposed only to the app's scoped asset protocol.
-
-## Evidence
-
-- Confirmed by code: relative links currently call `read_project_file`, which always decodes bytes as UTF-8.
-- Confirmed by code: attachment previews already use Tauri's scoped asset protocol and the existing media dialog.
-- Confirmed by the reported session: QA artifact links use `file:///tmp/...`, which the project-path normalizer intentionally rejects and previously reduced to plain labels.
-- Confirmed by tests: project paths are normalized and constrained to the active workspace.

package/external/pi-tools-suite/docs/browser-qa-subagent.md

@@ -1,177 +0,0 @@
-# Browser QA sub-agent specification
-
-## Goal
-
-Provide a cheap, fast `browser-qa` async-subagent that reproduces browser bugs
-and proves fixes with deterministic assertions plus screenshot, video, and trace
-evidence. Its ranked `models` list prefers `zai/glm-5.3-flash`, then
-`openai-codex/gpt-5.6-luna`, filtered by the active preset's model pool and
-confirmed runtime image support.
-
-## Inline agent workflow and skill isolation
-
-- All operating instructions, flow contracts, scenario-design guidance, and
-  auth-scaffolding rules live in `src/async-subagents/agents/browser-qa.md`.
-  Its body becomes the QA child's `promptAppend` through the shared agent
-  loader. Parent and router catalogs include only its short `description`.
-- Runner code, vendor dependencies/licenses, and optional JSONC examples live
-  under `src/async-subagents/agents/browser-qa/`; none is a discoverable skill.
-- Sub-agent processes disable normal extension discovery, then always load the
-  suite's model-tools extension. They load the Antigravity provider extension
-  only when an Antigravity model is explicitly selected.
-- A type profile may declare `isolatedSkills`. Spawning that profile adds
-  `--no-skills` followed by one explicit `--skill` per configured path.
-- `browser-qa` always disables normal skill discovery and filters skill flags
-  out of `extraArgs`, even without configured skills. It no longer injects a
-  mandatory QA skill. Explicitly configured skills are optional additions.
-- The launcher sets `PI_BROWSER_QA_RUNNER` to the absolute installed runner path,
-  replacing inherited values, and strips it from ordinary child environments.
-  QA invokes `node "$PI_BROWSER_QA_RUNNER"` without guessing paths from cwd.
-- Model-only profile overrides inherit the workflow. An explicit profile
-  `promptAppend` replaces the body like any other agent profile; it is not an
-  immutable security boundary. Runtime protections remain in the runner.
-
-## Authentication contract
-
-- Public browser QA requires no auth profile and does not create or require
-  `.pi/qa_auth.jsonc`. Its explicit base URL supplies the one exact allowed
-  origin, and the runner still blocks every other HTTP(S)/WebSocket origin.
-- Auth profiles live in project-local `.pi/qa_auth.jsonc` and are selected by
-  explicit id. The file must be a real project-local file with mode `0600` on
-  POSIX. Profile listings expose only `id`, description, and traits.
-- Listing profiles when the file is absent returns an empty list without side
-  effects. When authenticated QA explicitly requests credentials and that file
-  is absent, the runner creates a private empty template and returns
-  `provide_credentials`.
-  The sub-agent must explicitly ask the user to fill the reported file and
-  rerun QA; it must not read or edit the credential values itself.
-- Every profile requires one or more exact `allowedOrigins`. Secret-bearing auth
-  is applied only to those origins; all other HTTP(S)/WebSocket traffic and
-  service workers are blocked during QA.
-- Supported auth types are `form`, `cookie`, `localStorage`, `sessionStorage`,
-  `bearer`, and existing Playwright `storageState`.
-- The bundled runner reads secrets internally. Credentials must never be copied
-  into prompts, generated QA flows, shell arguments, transcripts, reports,
-  or QA evidence.
-- Generated browser state is private cache under `.pi/qa-auth-state`. Ephemeral
-  flows, evidence, and result manifests are written under the owning agent's
-  `.pi/subagents/<run>/<agent-id>/browser-qa/` workspace. Multiple profiles use
-  separate browser contexts/evidence directories, and normal sub-agent shutdown
-  or cleanup removes the whole workspace with its run.
-- Missing, rejected, or expired explicitly selected auth returns a
-  machine-readable update-required status naming only the profile id, config
-  file, and redacted reason. The parent asks the user to update the file and
-  reruns; there is no `/qa-auth` command.
-
-## QA execution contract
-
-- A model-authored QA flow is declarative JSONC, not executable JavaScript. The
-  trusted runner implements a bounded set of navigation, interaction,
-  assertion, screenshot, and auth-rejection actions and never gives the flow a
-  Playwright context or credential values.
-- Target discovery is a bounded preflight, not an open-ended research task. The
-  sub-agent invokes the runner within 45 seconds or returns `BLOCKED`; it does
-  not spend the full launcher budget reading source or probing prerequisites.
-- The launcher injects `PI_SUBAGENT_AGENT_DIR`, pre-creates a private
-  `browser-qa/flows/` workspace, and clears stale browser QA files when an agent
-  id is reused. The runner validates the directory's project/type metadata and
-  refuses flows outside it; the model cannot select a shared evidence root.
-- The runner owns browser lifecycle, origin checks, auth application, tracing,
-  screenshots, video finalization, and redacted result output. Before retaining
-  a trace it removes network/non-image resource entries, redacts configured and
-  runtime storage credentials, and verifies those values are absent.
-- After every navigation or visible interaction, the runner waits for DOM
-  readiness, completion of requests started by the action, and disappearance of
-  common visible busy/spinner/skeleton markers. It requires a 500 ms stable
-  interval before the next action so recordings remain readable; a page that
-  stays busy through the flow timeout fails closed instead of being tested as a
-  loading shell. App-specific readiness still requires an explicit declarative
-  wait/assertion in the authored flow.
-- Before any page is created, the runner installs a context-wide, isolated
-  interaction visualizer. Recorded clicks/double-clicks show a transient cursor
-  and pulse. Native drag/drop is replayed for 450 ms with a large orange cursor,
-  progressively drawn high-contrast path, and green drop marker. It also covers
-  same-origin frames, declared popups, and form-auth submission. The layer is
-  accessibility-hidden, pointer-transparent, never cancels application events,
-  and its bounded animations clear within the post-action stable interval.
-- Success and post-launch failure results include typed artifact groups. Every
-  item has an absolute filesystem path and a `file:` URI; the sub-agent must
-  present each item as a clickable Markdown link instead of reporting only the
-  evidence directory.
-- Success requires deterministic assertions. Visual inspection supplements,
-  but never replaces, explicit expected-state checks.
-- Auth rejection discovered by a QA flow is reported through the
-  `authRejectedIf` action so the parent gets an update-required status.
-
-## Reliability and shutdown contract
-
-- The built-in `browser-qa` profile has a 300-second wall-clock budget unless
-  the caller explicitly supplies a task or spawn timeout. This bounds model
-  stalls as well as browser work.
-- The trusted runner has its own bounded lifecycle. Browser launch, context
-  setup, auth, flow execution, evidence finalization, and browser shutdown must
-  not wait forever; a timeout reports the last started stage without exposing
-  flow contents or credentials.
-- Trace sanitization runs in a memory-limited worker that can be terminated at
-  the cleanup deadline; synchronous archive work cannot defeat the watchdog.
-- The launcher always writes a small sanitized `progress.jsonl` journal in the
-  agent directory. It records lifecycle/RPC event types and tool names, but not
-  prompts, tool arguments, tool results, model text, or secrets. The browser
-  runner writes similarly sanitized stage entries under its private workspace.
-- On POSIX, newly launched agents own a process group. Settled, timed-out, and
-  explicitly stopped agents signal that group rather than only the Pi process;
-  timeout/settled shutdown escalates to `SIGKILL` after its grace period. On
-  Windows the existing recursive `taskkill /T /F` behavior remains in force.
-- Process-tree cleanup is scoped to a launcher-created process-group marker so
-  an old or externally-created PID is never treated as an owned process group.
-  User browser sessions outside that group must not be signalled.
-- Playwright can launch Chromium in its own POSIX process group. On runner
-  failure the runner snapshots and kills only its own descendants before it
-  exits, covering that detached browser tree without touching a user's browser.
-
-## Acceptance criteria
-
-1. `browser-qa` resolves to the intended model/fallback and its inline Markdown
-   workflow, and its isolated child process can register the configured
-   model provider.
-2. Default QA spawn args contain `--no-skills` but no `--skill`. The child
-   receives the full workflow in its initial prompt and can invoke the bundled
-   runner through `PI_BROWSER_QA_RUNNER` from an unrelated project directory.
-   Optional configured skills still load; ordinary profiles retain existing
-   skill discovery behavior and do not receive QA-only environment paths.
-3. Auth profile listing and all error output are redacted; model-authored input
-   cannot execute code in the credential-bearing process.
-4. Runner tests cover public execution without an auth file, explicit profile
-   selection, all auth modes, fail-closed origins, path/mode hardening, private
-   empty-template creation only on an explicit auth request, non-executable
-   flows, and successful redacted evidence creation.
-5. Browser QA flows/evidence live only inside the owning sub-agent directory;
-   deleting the run removes them while persistent auth config/state remains.
-6. Runner tests prove that network activity and visible loading indicators are
-   awaited, persistent loading fails the flow, visible actions retain a stable
-   500 ms video interval, and context-wide click/drag video visualization is
-   installed with bounded click pacing.
-7. Completed test runs report clickable screenshot, video, and trace links
-   whenever those artifacts exist.
-8. Timeout tests identify the last browser stage, launcher progress remains
-   available when full RPC logging is disabled, and process-tree tests prove a
-   descendant is terminated without signalling unrelated processes.
-9. Suite tests/typecheck, host checks, and suite sync pass.
-
-## Real-browser regression test
-
-The repository includes a local mock-page E2E that launches real Chromium and
-asserts PNG screenshots, WebM video, sanitized trace output, and absolute
-path/`file:` URI metadata:
-
-```bash
-npx playwright install chromium
-npm run test:browser-qa-e2e
-```
-
-Normal suite tests keep this case skipped; the Publish workflow runs it on
-Linux after installing Chromium. The runner writes into a temporary simulated
-sub-agent directory. For manual inspection only, explicit E2E runs copy the
-latest artifacts to `.pi/qa-runs/browser-qa-e2e/latest/` and print clickable
-links; this test-only published copy is not the runtime storage contract. Set
-`BROWSER_QA_KEEP_EVIDENCE=0` to skip that copy.