Add watch plugin: /watch real-time Chrome monitor#2
Open
ShawnPana wants to merge 2 commits into
Open
Conversation
Colocated plugin (watch/) that monitors the user's Chrome live — network, console logs, and user actions (clicks/typing/navigations) plus a dense screen recording — and answers what happened on demand. Follows the active tab across all tabs. Depends on the browser-harness plugin + CLI, checked in the skill. - watch/.claude-plugin/plugin.json — plugin manifest - watch/skills/watch/SKILL.md — /watch entry; spawn monitor, stop+answer flow - watch/skills/watch/monitor.py — background CDP monitor (frames + timeline) - marketplace.json — register "watch" with source: ./watch - README — add watch row + install line Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
1 issue found across 6 files
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="watch/skills/watch/monitor.py">
<violation number="1" location="watch/skills/watch/monitor.py:76">
P2: Monitor can drop final browser events on shutdown because STOP is checked before a final event drain. This can miss the latest actions/network records right when answering a question.</violation>
</file>
Reply with feedback, questions, or to request a fix.
Fix all with cubic | Re-trigger cubic
| return sid | ||
|
|
||
| rec("watch.started", {"dir": WATCH_DIR}) | ||
| while not os.path.exists(STOP): |
There was a problem hiding this comment.
P2: Monitor can drop final browser events on shutdown because STOP is checked before a final event drain. This can miss the latest actions/network records right when answering a question.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At watch/skills/watch/monitor.py, line 76:
<comment>Monitor can drop final browser events on shutdown because STOP is checked before a final event drain. This can miss the latest actions/network records right when answering a question.</comment>
<file context>
@@ -0,0 +1,136 @@
+ return sid
+
+rec("watch.started", {"dir": WATCH_DIR})
+while not os.path.exists(STOP):
+ now = time.time()
+ tgts = page_targets()
</file context>
- monitor.py: purge old /tmp/watch-* dirs on start (keep newest 3) and switch the in-session frame trim to a time-based sweep so it isn't skipped when loop iterations are slow (many tabs) — fixes unbounded frame growth (a single session had ballooned to 2.4GB). - monitor.py: capture requestId->url so net.fail names the failing domain (e.g. ERR_NAME_NOT_RESOLVED -> https://dead-host/...). - SKILL.md: document the auto-cleanup + how to drop the current session. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
1 issue found across 2 files (changes from recent commits).
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="watch/skills/watch/monitor.py">
<violation number="1" location="watch/skills/watch/monitor.py:32">
P2: Session purge can delete active monitor directories from concurrent /watch runs. Restrict deletion to finished sessions (for example dirs already containing STOP) before rmtree.</violation>
</file>
Reply with feedback, questions, or to request a fix.
Fix all with cubic | Re-trigger cubic
| except OSError: pass | ||
| for _, p in sorted(dirs, reverse=True)[KEEP_DIRS:]: | ||
| if p != WATCH_DIR: | ||
| shutil.rmtree(p, ignore_errors=True) |
There was a problem hiding this comment.
P2: Session purge can delete active monitor directories from concurrent /watch runs. Restrict deletion to finished sessions (for example dirs already containing STOP) before rmtree.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At watch/skills/watch/monitor.py, line 32:
<comment>Session purge can delete active monitor directories from concurrent /watch runs. Restrict deletion to finished sessions (for example dirs already containing STOP) before rmtree.</comment>
<file context>
@@ -13,7 +13,25 @@
+ except OSError: pass
+ for _, p in sorted(dirs, reverse=True)[KEEP_DIRS:]:
+ if p != WATCH_DIR:
+ shutil.rmtree(p, ignore_errors=True)
+
+purge_old_sessions()
</file context>
Suggested change
| shutil.rmtree(p, ignore_errors=True) | |
| if os.path.exists(os.path.join(p, "STOP")): | |
| shutil.rmtree(p, ignore_errors=True) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Adds a
watchplugin, colocated as a subdirectory (watch/) withsource: ./watch— same pattern asqa./watchruns a background monitor over your Chrome and, when you ask, tells you exactly what happened. It captures two streams:It follows the active tab across all tabs (including tabs already open before you started) and is not a daemon you maintain — the agent spawns the monitor and stops it within the session.
Layout
How it works
On invoke, the skill spawns
monitor.pyviabrowser-harnessin the background. The monitor uses CDP auto-attach to follow every tab, captures actions via a nav-proofRuntime.addBindingchannel (with a JS-buffer fallback for pre-existing pages), pulls network/console from the event stream, and screenshots whichever tab is in front. On your next question it stops cleanly (aSTOPsentinel), reads the full timeline first, then loads only the relevant frame slice for visuals.Dependency on browser-harness
watchruns entirely through the browser-harness plugin + CLI. As withqa, there's no manifest-level "requires" field, so the dependency is documented in the description/plugin.jsonand theSKILL.mdverifiesbrowser-harnessis onPATHbefore running.Privacy & scope
Install (after merge)
Validation
marketplace.json(now 3 plugins) andplugin.jsonparse as valid JSON;monitor.pycompiles.🤖 Generated with Claude Code
Summary by cubic
Adds the
/watchplugin to monitor Chrome in real time and answer “what happened” with a structured timeline plus lightweight frame slices. Registerswatchin the marketplace and README, with safeguards to keep disk usage bounded.New Features
watch/plugin withplugin.json,skills/watch/SKILL.md, andskills/watch/monitor.py.browser-harness(no OS screen recording).Bug Fixes
/tmp/watch-*dirs (keep newest 3) and switch to time-based frame trimming to prevent runaway disk usage.net.failnow includes the failing request’s URL/domain by trackingrequestId -> url.Written for commit f0268c6. Summary will update on new commits.