@phnx-labs/agents-cli 1.18.3 → 1.18.4

package/CHANGELOG.md

@@ -1,5 +1,83 @@
 # Changelog
 
+## 1.18.4
+
+**Browser**
+
+- New canonical shape for action commands: bind the task once per shell via
+  `AGENTS_BROWSER_TASK`, then drop the positional `<task>` from every later
+  call. The old positional shape still works but emits a one-line deprecation
+  warning on stderr — agents and scripts can migrate at their own pace.
+  ```bash
+  export AGENTS_BROWSER_TASK=$(agents browser start --profile work)
+  agents browser navigate https://example.com
+  agents browser click 42
+  agents browser screenshot
+  ```
+  Env vars are per-process, so parallel agents in different shells never collide.
+  `--task <name>` is supported on every action command as an explicit one-off override.
+- `agents browser start` writes the resolved task name to **stdout** as a
+  single line (e.g. `swift-crab-falcon-a3f92b1c`), and routes the human
+  commentary ("Started task ... with tab ...", "Tip: export
+  AGENTS_BROWSER_TASK=...") to **stderr**. This makes
+  `T=$(agents browser start --profile X)` Just Work — no `--quiet` flag needed.
+- Auto-generated task names are now three English words plus an 8-char hex
+  suffix, e.g. `swift-crab-falcon-a3f92b1c`. Memorable, distinct, 32 bits of
+  entropy so parallel agents never collide. Daemon retries on the (vanishingly
+  rare) name clash and rejects explicit `--task <name>` values that already
+  exist.
+- `agents browser start --profile <name>` now pre-validates the profile
+  locally before touching the daemon. Missing profile prints the list of
+  available profiles plus the create-command hint instead of a generic error.
+- `agents browser tab list` is now `agents browser tabs` (top-level), pairing
+  cleanly with `agents browser tab focus <id>`. The old `tab list` form is
+  removed.
+- `agents browser --help` is reorganized by mental model — *Session lifecycle*,
+  *Drive the page*, *Capture evidence* — instead of an alphabetical dump.
+  Rare commands stay under a trailing *Commands* section.
+- BREAKING: `agents browser profiles prime` and `agents browser profiles launch`
+  are removed. Both were thin duplicates of `start`. For first-run
+  onboarding, just `agents browser start --profile <name>` and complete the
+  interactive screens in the browser; the user-data-dir persists across
+  runs. The daemon's `launch-profile` IPC action is also gone.
+- Named endpoint presets per profile. One profile can now cover the local
+  and remote variants of the same app instead of forcing two parallel
+  profiles. YAML supports both the legacy `endpoints: [url]` shape and the
+  new map form:
+  ```yaml
+  name: rush
+  browser: custom
+  electron: true
+  endpoints:
+    local:
+      target: cdp://127.0.0.1:9223
+      binary: /Applications/Rush.app/Contents/MacOS/Rush
+    mac-mini:
+      target: ssh://mac-mini?port=9223
+      # no binary — daemon attaches only
+  defaultEndpoint: local
+  ```
+  `agents browser start --profile rush --endpoint mac-mini` picks a specific
+  preset; `--endpoint` falls back to `defaultEndpoint` or the first preset.
+  Pre-validated client-side so a typo doesn't waste an IPC round-trip.
+  Per-endpoint `binary` and `targetFilter` override the profile-level
+  fields. `agents browser profiles show` lists every preset, marks the
+  default, and shows per-endpoint overrides.
+- The daemon's runtime identity is now `<profile>@<endpoint>` so the same
+  profile can run at multiple endpoints concurrently without colliding on
+  pid/port files. `agents browser status` and `tasks` show the composite
+  name, so you can tell at a glance which variant a task is using.
+- `agents browser screenshot --quality raw` captures pixel-faithful PNG
+  (no downscale) for archived QA evidence. Default stays `compressed`
+  (JPEG, capped near 100 KB) for chat-injected screenshots.
+- New `agents browser record start` / `agents browser record stop`
+  recording verbs. Captures via CDP `Page.startScreencast`, pipes frames
+  into ffmpeg (image2pipe → libvpx-vp9) and writes a webm under
+  `sessions/<task>/recordings/`. Bounded three ways — `--fps` (default
+  5), `--duration` (hard cap, default 60s), `--max-mb` (default 25);
+  whichever fires first auto-finalizes the file. Requires ffmpeg on
+  PATH (`brew install ffmpeg`).
+
 ## 1.18.3
 
 **Plugins** ([#22](https://github.com/phnx-labs/agents-cli/issues/22))

package/README.md

@@ -304,13 +304,20 @@ Give agents access to a real browser — no relay extension, no cloud service, n
 # Create an isolated profile for automation
 agents browser profiles create work --browser chrome
 
-# Start a task, navigate, interact
-agents browser start --profile work --task login-flow --url https://app.example.com
-agents browser refs login-flow              # Get interactive element refs
-agents browser click login-flow 42          # Click element ref 42
-agents browser type login-flow 15 "hello"   # Type into element ref 15
-agents browser screenshot login-flow        # Smart resizing, token-efficient
-agents browser done login-flow              # Close task's tabs when finished
+# Start a task once, then bind it to this shell — every later command picks it up.
+# `start` writes the resolved name (e.g. `swift-crab-falcon-a3f92b1c`) to stdout
+# and human-friendly commentary to stderr, so $(...) capture stays clean.
+export AGENTS_BROWSER_TASK=$(agents browser start --profile work --url https://app.example.com)
+agents browser refs                  # Get interactive element refs
+agents browser click 42              # Click element ref 42
+agents browser type 15 "hello"       # Type into element ref 15
+agents browser screenshot            # Smart resizing, token-efficient
+agents browser tabs                  # List tabs open for the current task
+agents browser tab focus tab123      # Switch focus to another tab
+agents browser done                  # Close task's tabs when finished
+
+# Need to address a different task in the same shell? Override per call:
+agents browser screenshot --task other-flow
 ```
 
 ### Why this works where Playwright fails