Works with the AI Agents you already use
GitHub pull_request_target Block Fix (Dec 2, 2026): Audit, Migrate and Allow Workflows Safely
Find every workflow that stops running when GitHub blocks pull_request_target on December 2, 2026, fix each one with the safest pattern, and prove it with a scanner exit code.
$1
GitHub pull_request_target Block Fix (Dec 2, 2026): Audit, Migrate and Allow Workflows Safely
Example session with this skill installed
Our public repo had failed runs on Nov 9 about pull_request_target. Will our workflows break on December 2? Fix whatever needs fixing.
- Read your context and instructions
- Compiled the github pull_request_target block
pull_request_target audit — acme/widget — 2026-11-12
Deadline: 2026-12-02 (public repos, GitHub.com). Scanner exit before: 2, after: 1.
Inventory
| Workflow | Triggers | Secrets | Write token | Fetches fork code | Executes it | Verdict |
|---|---|---|---|---|---|---|
| labeler.yml | pull_request_target | no | pull-requests: write | no | no | B keep and allow |
| lint.yml | pull_request_target | no | no | no | yes (base branch) | A switch |
| preview.yml | pull_request_target | VERCEL_TOKEN | default (no block) | yes, L9 head.sha | yes, npm ci/build | D fix now |
Scanner findings (preview.yml): CRITICAL pwn-request; HIGH L9 fork-checkout; HIGH L13 script-injection (pull_request.title in run:); MEDIUM L7 checkout pinned to v4.1.1; MEDIUM no permissions block.
Changes
- preview.yml disabled first (gh workflow disable "Preview deploy"), then rebuilt as pr-build.yml (pull_request, contents: read, uploads dist/ + validated PR number) and pr-preview.yml (workflow_run, deploys the static files only, VERCEL_TOKEN via env:).
- lint.yml: trigger changed to pull_request, permissions: contents: read.
- labeler.yml: unchanged; already least-privilege.
Policies to create
| Policy | Scope | Workflow path | Event allowed |
|---|---|---|---|
| allow-prt-labeler | acme/widget | .github/workflows/labeler.yml | pull_request_target |
Verification
| Check | Result |
|---|---|
| V1 scanner: no CRITICAL/HIGH (exit 1, one INFO for labeler) | PASS |
| V2 policy scoped to labeler.yml | PENDING (owner must create) |
| V3 YAML parses | PASS |
| V4 required check names unchanged ("lint", "build") | PASS |
| V5 fork test PR | PENDING |
| V6 Policy insights | PENDING |
Connects securely to your tools. The creator never sees your data.
About this skill
On December 2, 2026 GitHub Actions blocks pull_request_target by default in every public repository on GitHub.com unless an Actions event policy explicitly allows it. Labelers, preview deploys, coverage commenters and AI review bots that run on fork pull requests will stop silently. The first brownout ran on November 9; the next is November 16, 16:00–18:00 UTC.
This skill gives your agent a complete, verified procedure:
- Inventory: a bundled, read-only, standard-library Python scanner lists every
pull_request_target,workflow_runandissue_commentworkflow, flags fork checkouts,gh pr checkout/git fetch pull/steps,allow-unsafe-pr-checkout, script injection, missingpermissions:and unsafeactions/checkoutpins. Exit codes 0/1/2 make it CI-friendly. - Classify: four verdicts (A switch to
pull_request, B keep and allow with a file-scoped event policy, C split intopull_request+workflow_run, D fix-now pwn request) from four yes/no questions. - Fix: copy-ready, YAML-validated patterns for each verdict, including a safe artifact hand-off that validates the PR number and never executes fork output, comment-bot guards and the
env:script-injection fix. - Allow: exact Settings → Actions → Policies steps and
gh apichecks for repository and organization policies. - Prove: a six-point PASS/FAIL verification table and a
prt-audit.mdreport with diffs.
Every rule is cited to GitHub's own changelog and docs, checked on 2026-11-12.
The scanner is line-based, not a full YAML parser, so the agent confirms each verdict by reading the file. It does not create Actions event policies for you (policy creation needs admin rights and is done in the GitHub UI). It covers GitHub.com public repositories; GitHub Enterprise Server and private/internal repositories are out of scope because the December 2 rule does not apply to them. Facts were checked on 2026-11-12; if GitHub changes the dates again, re-check the changelog.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 5 days ago
Requires
Permissions
File Scopes
The bundled scanner is read-only and makes no network calls. Optional gh commands contact api.github.com with the user's own GitHub CLI login.