Repository navigation
Add Agent Session Board canvas 🤖🤖🤖 - #6491
Conversation
…ry work Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
🔒 PR Risk Scan ResultsScanned 19 changed file(s).
Skipped non-text or missing files
|
🧩 Canvas smoke test✅ Passed · 0 error(s), 4 warning(s)
Plugin
|
| Field | Value |
|---|---|
| Description | Supervise several Copilot sessions working one GitHub repository: a decision queue, progress by label track, merge-gate status for every PR, hourly and daily merge flow, warnings, and an optional merge-lock lane. |
| Author | hellices — https://github.com/hellices |
| License | MIT |
| Keywords | canvas, copilot-extension, dashboard, merge-queue, multi-agent, pull-requests, supervision |
| Extensions | agent-session-board |
| Materialize | ✅ |
| Install | ✅ |
Changed plugin files
plugins/agent-session-board/README.mdplugins/agent-session-board/plugin.json
Extension agent-session-board
| Field | Value |
|---|---|
| Registered by | agent-session-board |
| Package | agent-session-board 1.0.0 |
| Preview | 1280×720, 158.2 KB |
| Modules | 14 (8 reachable from extension.mjs) |
| Files | 17 |
Permissions and capabilities (static analysis of modules reachable from extension.mjs)
- Spawns processes (node:child_process)
- File system access (node:fs)
- Network sockets / local server (node:http, node:net, ...)
- Worker threads / child processes (node:worker_threads, node:cluster)
- Host information (node:os)
- Outbound HTTP requests (fetch)
- Reads environment variables (process.env)
Warnings
- test/collect.test.mjs: import "node:test" is not a Node.js builtin, host-provided package, or declared dependency (module is not reachable from extension.mjs)
- test/exec.test.mjs: import "node:test" is not a Node.js builtin, host-provided package, or declared dependency (module is not reachable from extension.mjs)
- test/pattern.test.mjs: import "node:test" is not a Node.js builtin, host-provided package, or declared dependency (module is not reachable from extension.mjs)
- test/store.test.mjs: import "node:test" is not a Node.js builtin, host-provided package, or declared dependency (module is not reachable from extension.mjs)
Changed extension files
extensions/agent-session-board/README.mdextensions/agent-session-board/assets/preview.pngextensions/agent-session-board/extension.mjsextensions/agent-session-board/lib/collect.mjsextensions/agent-session-board/lib/demo.mjsextensions/agent-session-board/lib/exec.mjsextensions/agent-session-board/lib/match-worker.mjsextensions/agent-session-board/lib/page.mjsextensions/agent-session-board/lib/pattern.mjsextensions/agent-session-board/lib/server.mjsextensions/agent-session-board/lib/store.mjsextensions/agent-session-board/package.jsonextensions/agent-session-board/test/collect.test.mjsextensions/agent-session-board/test/exec.test.mjsextensions/agent-session-board/test/fixtures/board-writer.mjsextensions/agent-session-board/test/pattern.test.mjsextensions/agent-session-board/test/store.test.mjs
Static checks compile modules without executing them. Minimum preview size is configurable with CANVAS_PREVIEW_MIN_WIDTH/CANVAS_PREVIEW_MIN_HEIGHT. Workflow run
🚦 Submission status: 👀 Ready for reviewRisk tier: Why this tier
Automated checks
Review
Commands
Updated for |
There was a problem hiding this comment.
🟡 Changes recommended
Unresolved critical flaws allow unauthorized access and decision submission, lost updates across sessions, and false merge-readiness signals from unknown or incomplete GitHub data.
16 open issues
Treat unknown mergeability as a pending gate · New Restrict checklist indentation matching to spaces and tabs · New Paginate open PR collection before deriving lock state · New Paginate review threads before declaring PRs mergeable · New Prevent untrusted sites from framing decision controls · New Authenticate local clients before exposing state or accepting answers · New Add inter-process locking around shared board updates · New Calculate lock expiry dates in the supplied timezone · New Verify comment authors before applying lock events · New Align chart buckets with rolling reporting intervals · New Check answer IDs with an own-property test · New Report GraphQL errors alongside usable partial data · New Avoid presenting truncated merge searches as complete totals · New Paginate track issues before calculating totals and readiness · New Paginate decision-label search to avoid hiding questions · New Preserve answer drafts and focus during automatic refresh · New
What changed in this PR
Adds Fleet Control Board, a canvas extension and installable plugin for supervising multiple Copilot sessions working in one GitHub repository.
Changes:
- Adds decision queues, track progress, session merge gates, merge-flow charts, warnings, and optional merge-lock tracking.
- Implements GitHub CLI data collection, persistent board state, a local server, and an English/Korean interface.
- Includes demo data, focused unit tests, usage documentation, and plugin catalog registration.
| File | Description |
|---|---|
plugins/fleet-control-board/README.md |
Documents plugin installation and usage. |
plugins/fleet-control-board/plugin.json |
Registers the plugin and bundled extension. |
extensions/fleet-control-board/test/store.test.mjs |
Tests storage and input validation. |
extensions/fleet-control-board/test/collect.test.mjs |
Tests attribution, locks, gates, and state assembly. |
extensions/fleet-control-board/README.md |
Documents features, actions, configuration, and privacy. |
extensions/fleet-control-board/package.json |
Defines extension metadata and the test command. |
extensions/fleet-control-board/lib/store.mjs |
Provides persistent state and input validation. |
extensions/fleet-control-board/lib/server.mjs |
Serves the board and local API endpoints. |
extensions/fleet-control-board/lib/page.mjs |
Implements the localized dashboard and answer controls. |
extensions/fleet-control-board/lib/demo.mjs |
Supplies fictional demonstration data. |
extensions/fleet-control-board/lib/collect.mjs |
Collects GitHub data and derives board metrics. |
extensions/fleet-control-board/extension.mjs |
Connects canvas lifecycle, actions, and caching. |
docs/README.plugins.md |
Adds the plugin to the documentation catalog. |
.github/plugin/marketplace.json |
Adds the marketplace entry. |
Review effort: Balanced
…nglish default Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
🟡 Changes recommended
Stale-lock recovery can delete another writer’s live lock and lose board updates, while excessive regular-expression backtracking can freeze all boards and actions.
8 open issues
Prevent configured regexes from blocking the event loop · New Make stale-lock recovery and cleanup ownership-aware · New Reject answers submitted for replaced questions · New Match answer revision when acknowledging a decision · New Verify repository write permission before allowing lock actions · New Distinguish repeated local hours during DST aggregation · New Expose PRs beyond the first eight · New Provide accessible data tables for both charts · New
16 resolved since last review
Add inter-process locking around shared board updates Authenticate local clients before exposing state or accepting answers Prevent untrusted sites from framing decision controls Paginate review threads before declaring PRs mergeable Paginate open PR collection before deriving lock state Restrict checklist indentation matching to spaces and tabs Treat unknown mergeability as a pending gate Preserve answer drafts and focus during automatic refresh Paginate decision-label search to avoid hiding questions Paginate track issues before calculating totals and readiness Avoid presenting truncated merge searches as complete totals Report GraphQL errors alongside usable partial data Check answer IDs with an own-property test Align chart buckets with rolling reporting intervals Verify comment authors before applying lock events Calculate lock expiry dates in the supplied timezone
Review effort: Balanced
…visions, DST buckets, accessibility Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
🟡 Changes recommended
Unresolved critical issues can freeze the board during pattern matching, falsely report PRs as merge-ready, and leave earlier decision answers valid after question bodies change.
9 open issues
Mergeable status ignores outstanding repository merge requirements · New Decision fingerprints omit label-raised question body changes · New Process command matching bypasses the shared subject-length cap · New Character-class scanner misparses an initial closing bracket · New Overlapping quantifiers can still cause catastrophic backtracking · New Per-PR reference truncation hides incomplete attribution · New Prototype-named labels corrupt count and color dictionaries · New Unpaginated closing-issue lookup silently omits linked issues · New Light-theme status and link colors fail contrast requirements · New
Review effort: Balanced
…hing, attribution limits, light theme contrast Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
🟡 Changes recommended
Appending at capacity silently deletes unanswered questions, while truncated revision hashes could allow answers to be reused after contributor-controlled question changes.
5 open issues
9 resolved since last review
Overlapping quantifiers can still cause catastrophic backtracking Character-class scanner misparses an initial closing bracket Process command matching bypasses the shared subject-length cap Decision fingerprints omit label-raised question body changes Mergeable status ignores outstanding repository merge requirements Light-theme status and link colors fail contrast requirements Unpaginated closing-issue lookup silently omits linked issues Prototype-named labels corrupt count and color dictionaries Per-PR reference truncation hides incomplete attribution
Review effort: Balanced
…est, permission level, lock backoff Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
🔵 Needs a closer look
Human assessment is needed because background refreshes can bind an existing draft to a changed question, allowing an answer intended for an earlier revision to be submitted.
No open issues 🎉
5 resolved since last review
Review effort: Balanced
… and use worker threads in tests Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
There was a problem hiding this comment.
🔵 Needs a closer look
Human assessment is needed because unpublished findings identify critical risks of submitting drafts against changed questions and reporting PRs as mergeable from incomplete review-thread data.
No open issues 🎉
Review effort: Balanced
|
/request-review |
|
🙋 Added |
aaronpowell
left a comment
There was a problem hiding this comment.
Thanks for contributing this canvas. It adds a focused workflow for supervising sessions across PR gates, label-based progress, and question handoffs. The repository’s existing Agent Radar is focused on local session and data-flow visibility, while this board is centered on work moving through a configured repository. CI reports the plugin validation, build, and canvas smoke test passing, and the README documents the gh-based data path and demo mode. This is a useful addition to the Copilot canvas collection.



Pull Request Checklist
npm startand verified thatREADME.mdis up to date.mainbranch for this pull request.Description
Agent Session Board is a canvas for checking on several Copilot agent sessions and managing them while they work on one GitHub repository through issues, pull requests, reviews and merges.
Value: when several sessions run at once, the person supervising them needs to know three things at any moment:
This board answers all three on one screen and keeps them current. It also gives agents a two-way channel back to that person.
Waiting on you: agents post questions with
set_decisions. The user answers on the board, and agents read the answers withget_answersand acknowledge them withmark_answers_seen. Open issues that carry a configurable decision label are added too.Progress by track: one row per configured label (milestone, area, phase), with checklist progress, closed and total issues, and merge counts. Optional bars show epic sub-issue progress. Each merge is assigned to a track by checking, in order:
Sessions: one card per agent session, matched by PR number, branch pattern or title pattern. Each PR shows the first merge gate it still has to pass, one of:
Idle sessions are flagged.
Merge flow: hourly bars for the last 24 clock hours and daily bars for the last 7 calendar days, stacked by track. Each chart has a data table for screen-reader and keyboard users.
Warnings: CI failures, conflicts, stale PRs, a low API budget, expired merge locks, and data too large to load completely.
Local resources (optional): the header shows free memory, swap and heavy-job counts, with limits.
Merge lock lane (optional): reads
merge lock:,merge reserve:andmerge lock release:lines from a coordination issue. A line counts only when its author's write access is verified, or when the author is on a configuredlockAuthorslist.Open it with
{ "repo": "owner/name" }, or with{ "demo": true }for fictional sample data. Its actions arerefresh,get_summary,configure_board,set_decisions,get_answersandmark_answers_seen. The board is in English by default. Callingconfigure_boardwithlanguage: "ko"switches it to Korean.Safety and data handling
ghCLI. The board never reads or stores tokens.lib/exec.mjs. It runs only an allowlist:gh, plusps,sysctlandmemory_pressurefor the optional local checks. Commands run without a shell, with arguments as an array, a timeout and an output cap.ghis missing or signed out, the board shows the setup steps instead of loading data. Demo mode works withoutgh.127.0.0.1on an ephemeral port.<copilot-home>/extension-data/agent-session-board/. Each file has a format marker and version, and is written atomically under a lock file, so several sessions can share one board safely.open()is idempotent perinstanceId.The screenshot uses the built-in demo data only.
Type of Contribution
Additional Notes
This PR was authored by an AI agent (GitHub Copilot) for @hellices, per the CONTRIBUTING note on AI-authored submissions. A human reviews it before it leaves draft.
Validation
node --checkon every module: passed.node --test test/*.test.mjsinextensions/agent-session-board: 36/36 passed. The tests cover:npm run plugin:validateandnpm run build: passed.extensions_reload, thenlist_canvas_capabilities.open_canvaswith demo data and withgithub/awesome-copilot.refresh,get_summaryandconfigure_board.set_decisions, an answer from the page,get_answers, thenmark_answers_seen.instanceIdreturned the same URL, and settings came back after an extension restart.By submitting this pull request, I confirm that my contribution abides by the Code of Conduct and will be licensed under the MIT License.