create-pr

Open GitCode merge requests with local CI validation and result interpretation.

64|15|Updated Aug 14, 2026
One-click install
npx skills add https://github.com/openJiuwen-ai/sciencediscovery --skill create-pr-openjiuwen-ai
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: create-pr
Source: https://github.com/openJiuwen-ai/sciencediscovery/tree/main/.agents/skills/create-pr
Command: npx skills add https://github.com/openJiuwen-ai/sciencediscovery --skill create-pr-openjiuwen-ai

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Opening a merge request for the ScienceDiscovery project involves strict gates: running the UT, ST, and E2E test layers locally, pushing only to a personal GitCode fork, writing a reviewable body with real test numbers, and correctly interpreting bot comments and CodeArts pipeline results. This Skill encodes that entire process so nothing is skipped or misread. ## Core Features & Use Cases - Pre-merge validation gate: Runs pnpm ci:ut, ci:st, and ci:e2e on the exact commit to be pushed, with sandbox checks and clean-worktree guidance, and attributes failures against an unmodified origin/main baseline. - Correct GitCode targeting: Pushes task branches only to the operator's own GitCode fork and creates the merge request against openJiuwen/sciencediscovery with --head <login>:<branch> --base main, avoiding the GitHub mirror. - Result interpretation: Explains the openJiuwen-bot and atomgit-bot comments, the CodeArts result table (SCA, UT, ST, binary jobs), ci-* labels, and how to read logs via gitcode pr view and the REST API. - Use Case: After finishing a backend change to Runs and artifacts, run all three CI layers locally, collect the pass/skip counts, open the merge request from your fork with a body documenting the E2E journeys, then read back the bot's pipeline result to confirm ci-successful. ## Quick Start Ask the assistant to create a merge request for the current branch on GitCode, running the UT, ST, and E2E layers first and reporting the results in the body.

Frequently Asked Questions about create-pr

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a merge request on GitCode for ScienceDiscovery?▼

Run the UT, ST, and E2E layers locally on the commit to push, push the branch only to your own GitCode fork, then create the merge request with --head <login>:<branch> --base main targeting openJiuwen/sciencediscovery. Read the request back with gitcode pr view to confirm it landed correctly.

How do I run the UT, ST, and E2E test layers before opening a PR?▼

Verify the sandbox with bwrap, then run pnpm ci:ut, pnpm ci:st, and pnpm ci:e2e with CI_RESULTS_DIR and CI_RUNTIME_DIR set. Run them on the exact commit being pushed, using a detached clean worktree if the working tree has other changes.

Why should I not push my branch to origin or open a PR on GitHub?▼

GitHub is only a synced mirror with separate SHAs, so a merge request opened there is in the wrong place. Task branches must live on your own GitCode fork, and the merge request targets openJiuwen/sciencediscovery on gitcode.com.

What do the ci-running, ci-successful, and ci-failed labels on a merge request mean?▼

The openJiuwen-bot adds ci-running while the CodeArts pipeline executes, then ci-successful or ci-failed when it finishes. Judge the run by its latest result comment rather than the label set, since ci-running may linger after the result posts.

What should I do when a CI layer fails before opening a merge request?▼

Attribute the failure first by running the same layer on unmodified origin/main in a separate worktree. An identical failure is pre-existing and should be stated in the body; a failure only on your commit must be fixed and the layer rerun before opening the request.

Can I skip the E2E section for a backend-only change?▼

No. Backend-only changes to Runs, tools, versioned state, artifacts, or permissions still require E2E journeys through the API, CLI, or local stack. Only pure documentation changes may mark E2E not applicable, with a concrete reason.