Skills Plugins MCP Prompt Model 博客 我的中心

pr-full-review

Review an AnkiDroid pull request or branch against AnkiDroid's conventions.

DeepseekModel Curated skill Quality Excellent · 90 v1.0.0

Get

https://deepseekmodel.com/api/download.php?id=ankidroid-anki-android-agents-skills-pr-full-review-skill-md&format=skill
Download .skill Standard format with system_prompt and model_config, ready for any agent framework
The actual content of the system_prompt field in the .skill file.
name pr-full-review description Review an AnkiDroid pull request or branch against AnkiDroid's conventions. AnkiDroid PR Full Review Review the PR against the standards in reviewer-guide.md . Report findings in three groups: blocking gates , spec conformance (does the PR do what its issue asked?), then non-blocking nit: s. Never act on GitHub. Output the review in this conversation. Do not post comments or reviews, approve, request changes, label, or merge. Skill Usage Accepts a PR number, a PR URL, or no argument (in which case, review the current branch against upstream/main ): pr-full-review 21206 pr-full-review https://github.com/ankidroid/Anki-Android/pull/21206 pr-full-review review the checked-out branch 's diff against `main` Fetching the PR Fetch the PR with the gh command. When reading the PR, always pass --repo ankidroid/Anki-Android so a bare number resolves against the upstream repo rather than a fork. Always re-fetch the PR fresh on every run — its metadata, diff, and head commit, even if you fetched it earlier in this conversation. PRs change between reviews; never reuse cached context. Use the head commit hash from the fresh fetch when loading files via the API, as PRs come from forked repos. Don't review from the diff hunks alone . Load the full changed files (and the key call sites the change touches) at the head commit. A hunk hides whether a change is correct in its surrounding context. Before reviewing Read the existing PR comments and review threads first. Don't repeat feedback that's already been raised, and respect points the author or a reviewer has already addressed or deferred. gh pr view <number> --repo ankidroid/Anki-Android --comments # conversation comments gh api repos/ankidroid/Anki-Android/pulls/<number>/comments # inline code-review comments Unaddressed maintainer requests are blocking. Read the full discussion on both the PR and the linked issue (see Spec conformance ). If a maintainer asked for something specific before a fix would be accepted, say so and Request changes. Spec conformance Report whether the PR does what it set out to do, as its own section . Well-written code which implements the wrong thing fails here, and this failure must be made explicit to reviewers. Determine the spec, in order: A linked issue - Fixes # , Closes # , Resolves # , or Part of # in the body or commits (typically written as Fixes <N> - the commit message may omit the # ). Also include the PR's own Purpose / Description . Open the issue and read it in full — including its comments . The decisive context often lives in the thread, not the description: gh issue view <number> --repo ankidroid/Anki-Android --comments Bug fixes: establish the root cause before endorsing the fix For a bug fix, spec conformance means the PR fixes the actual cause , not that it plausibly might. Verify the author's fix: Trace the real code path that triggers the bug (load the call sites and any resource/theme values they read), and confirm the claimed trigger can actually occur. If your trace shows it *can't occur the way the PR describes, then the root cause is not understood. Say so explicitly. A fix applied without an understood, demonstrated trigger is tech debt: . Treat "unknown root cause" as a blocking finding and ask for the diagnostics / reproduction needed to establish it first unless it is explicitly acknowledged as not being understood. State that it risks masking the real bug and the fallback behavior can't be validated against a scenario nobody has captured. Confirm the bug-fix commit actually evidences a reproduction. Output Open the review with a one-line verdict, mirroring GitHub's three review actions — the reader will pick one of these on the PR: Verdict: [Request changes / Comment / Approve]: Request changes: a blocking gate fails (red/pending CI, unfilled template, merge commits, missing license header, or a closing keyword on a partially-resolved issue), an explicit maintainer request was skipped, a bug fix lands without an established/verified root cause, or a correctness, spec, or test finding must be fixed before merge. Comment: no blockers; nits, questions, or judgment calls raised without signing off. Approve: gates pass and the change is sound - only nit: s remain, if anything. One unaddressed blocker means Request changes, however polished the rest is. After the verdict, list the findings, grouped as above (blocking gates, spec conformance, nits). Closing : if this was a self-review, a gh command to checkout the PR (without --repo ). Close with a final line linking the PR which was reviewed.
Keywords that activate this skill. Click one to copy it.

This skill does not provide trigger words.

The downloaded .skill package contains the following fields.
Field Description
formatFormat tag (skill/v1)
skill_idUnique skill ID
nameSkill name
versionVersion
descriptionDescription
categoryCategories (array)
trigger_wordsTrigger words
tagsTags
sourceSource
source_urlSource URL (this page)
exported_atExported at (set per download)
system_promptSystem prompt body
model_configModel config: provider / model / temperature / max_tokens / top_p
examplesExamples
install_guideImport guide for Coze / Dify / Claude / custom frameworks
The same skill can be exported in different platform formats.
.skill Standard format with system_prompt and model_config, ready for any agent framework Download
.skillpro Enhanced format with scripts, tools, dependencies and hooks Download
.json Plain JSON export with system_prompt and model parameters only Download
Coze Markdown with frontmatter, for Coze platform import Download
Dify Dify DSL, import directly after creating an app Download

每日精选 Skill 推荐,免费送到你邮箱

输入邮箱,每天接收一个精选 AI Agent 技能推荐。完全免费,持续更新。

提交后我们会发送一封确认邮件,点击邮件里的链接才会开始收信。

完全免费,取消任意时间。我们不会发送垃圾邮件。