New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Phyai Solve PR Comments

by mingti-org in mingti-org/phyai

Triage and resolve review comments on a GitHub PR — fetch all comment surfaces (issue / inline / review), validate each suggestion against upstream source rather than trusting blindly, present…

MITAuto-check passed

Install Phyai Solve PR Comments

skills CLI
$ npx skills add mingti-org/phyai --skill phyai-solve-pr-comments -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install mingti-org/phyai phyai-solve-pr-comments --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/mingti-org/phyai.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/phyai-solve-pr-comments .claude/skills/phyai-solve-pr-comments && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
phyai-solve-pr-comments
GitHub stars
129
Token cost
~1.9k tokens
SKILL.md length
840 words
Files
1
Skills in repo
5
Repo updated
First seen
Licence
MIT

At a glance

Triage and resolve review comments on a GitHub PR — fetch all comment surfaces (issue / inline / review), validate each suggestion against upstream source rather than trusting blindly, present…

  • Works in 7 steps: Fetch every comment surface — don't… → Triage before doing anything → Validate every claim against ground truth → …
  • The user asks to check PR Ns comments
  • SKILL.md covers When to use, Workflow, Anti-patterns and Quick reference: gh CLI cookbook
  • Calls gh, git and pytest

What it does

Phyai Solve PR Comments is an agent skill from mingti-org/phyai. Triage and resolve review comments on a GitHub PR — fetch all comment surfaces (issue / inline / review), validate each suggestion against upstream source rather than trusting blindly, present findings to the user before editing, then make focused changes and re-run tests. Use when the user asks to "check PR N's comments", "address review feedback", "resolve PR comments", or similar.

Its SKILL.md is about 1.9k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It works with GitHub. The repository describes itself as: PhyAI is a high-performance framework for running Physical AI models (VLA, WAM, and beyond), supporting both cloud-based serving and on-device deployment. The licence is MIT.

When your agent uses it

  • The user asks to check PR Ns comments
  • Address review feedback
  • Resolve PR comments

Example prompts

  • “check PR N”
  • “address review feedback”
  • “resolve PR comments”
  • “/phyai-solve-pr-comments”

Workflow steps

7 steps, taken from the step headings in SKILL.md.

  1. Fetch every comment surface — don't trust gh pr view --comments
  2. Triage before doing anything
  3. Validate every claim against ground truth
  4. Present findings before editing
  5. Make the changes — minimal and focused
  6. Verify before reporting done
  7. Reply on the PR (only if user asks)

What it can do on your machine

Read from SKILL.md and the folder at commit 36a46bf. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • gh
    • git
    • pytest

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use gh and git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Phyai Solve PR Comments loads about 1.9k tokens when it runs. Until then it costs about 103 tokens; SKILL.md has 840 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~103
When it runs · the whole SKILL.md, loaded when a task matches
~1.9k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from mingti-org/phyai at commit 36a46bf, republished under its MIT licence (© mingti-org). 840 words, ~1,908 tokens.

Download SKILL.mdSave it as .claude/skills/phyai-solve-pr-comments/SKILL.md (or your agent's skills folder).
name
phyai-solve-pr-comments
description
Triage and resolve review comments on a GitHub PR — fetch all comment surfaces (issue / inline / review), validate each suggestion against upstream source rather than trusting blindly, present findings to the user before editing, then make focused changes and re-run tests. Use when the user asks to "check PR N's comments", "address review feedback", "resolve PR comments", or similar.

Solve PR Comments

The point of this skill is to resolve PR review feedback correctly, not just compliantly. Bot suggestions (gemini-code-assist, copilot, codex) are often partially right or confidently wrong — they pattern-match local code without checking the upstream library contract. Your job is to verify each claim against ground truth, then surface the verdict to the user before touching code.

When to use

User says one of:

  • "check PR N's comments"
  • "address review feedback on PR N"
  • "resolve / fix PR N comments"
  • "look at the comments on PR N"
  • "there are comments on PR N, fix them"

Skip this skill for:

  • Asking what a comment says (just gh pr view)
  • Re-running existing review (use /review instead)

Workflow

1. Fetch every comment surface — don't trust gh pr view --comments

gh pr view N --comments quietly drops content when GitHub's GraphQL flags Projects-classic deprecation. Hit the REST endpoints directly:

bash
# Conversation comments (issue thread)
gh api repos/OWNER/REPO/issues/N/comments \
  --jq '.[] | {id, user: .user.login, body, created_at}'

# Inline review comments (the line-anchored ones)
gh api repos/OWNER/REPO/pulls/N/comments \
  --jq '.[] | {id, user: .user.login, path, line, side, body, created_at}'

# Overall reviews (the summary block on top of each review)
gh api repos/OWNER/REPO/pulls/N/reviews \
  --jq '.[] | {id, user: .user.login, state, body, submitted_at}'

If gh auth status says "token invalid" — stop and ask the user to re-auth. Don't try to work around it.

2. Triage before doing anything

For each comment, classify:

ClassExamplesAction
Real bug / perf issue"race condition in shared dict", "per-forward fp32 cast"Plan a fix, verify approach with user
Wrong claimBot extrapolated from a pattern that doesn't apply to this codeReject — explain to user why it's wrong (cite upstream source)
Doc-onlySubtle contract that's not visible from codeAdd docstring/comment, no code change
Style / nitNaming, formattingSkip unless user asks; not the point of review
Already fixedComment is on stale codeMark as outdated, move on
3. Validate every claim against ground truth

This is the highest-leverage step. Before believing a bot:

  • Library contract claims ("X needs fp32") -> read the upstream source. CUDA kernels usually live in .../site-packages/<pkg>/data/csrc/*.cu. Look for TORCH_CHECK / TVM_FFI_ICHECK / dispatch macros.

  • Reference implementations -> grep .tmp/<reference-project>/ (e.g. sglang, vllm, lerobot) for how mature codebases handle the same kernel. SGLang's python/sglang/srt/layers/ is a particularly good reference for kernel wrappers. If the repo you need isn't already under .tmp/, clone it — git clone --depth 1 <upstream-url> .tmp/<name> is enough for a read-only consult; full history is rarely needed. Don't try to reason from training-data memory of the source.

    STRICT — DO NOT COPY CODE FROM REFERENCE REPOS. They exist only for correctness verification: confirming a kernel's dtype contract, double-checking a math formula, comparing dispatch logic. The fix you write must be authored from scratch in this project's style and abstractions. Pasting an SGLang/vLLM/lerobot block — even one that "looks like it fits" — is a license violation, breaks our layer hierarchy, and drags in dependencies we don't want. Read their code, understand the constraint, then write our own.

  • Project tests -> read existing tests under tests/ before editing. They encode the contract the user already expects.

A bot's claim is worth nothing until you've checked it against the upstream contract. The bot may have pattern-matched a similar-looking issue from a different library; the right fix here may be the opposite of what they suggested, or there may be no fix at all (and the contract just needs to be documented).

Show full SKILL.md (331 more words)Show less
4. Present findings before editing

Reply to the user with a tight table:

| # | File:Line | Issue | Verdict | Plan |
|---|---|---|---|---|
| 1 | foo.py:100 | claim X | Real | Pre-allocate fp32 in __init__ |
| 2 | bar.py:50  | claim Y | Wrong | Bot misread; weight dtype must match input. Add docstring note instead. |
| 3 | baz.py:200 | claim Z | Stale | Already fixed in commit abc123 |

Wait for the user to confirm scope. Do not silently apply every bot suggestion.

5. Make the changes — minimal and focused
  • One concept per edit. Don't bundle "address comment + refactor + rename".
  • Update docstrings for any contract you discovered (e.g., "flashinfer requires weight dtype == input dtype"). Future readers shouldn't have to re-derive what you just learned.
  • If a bot suggestion has a code block (a suggestion: block in the comment body), still read it but adapt to local conventions; don't paste verbatim if it conflicts with project style.
6. Verify before reporting done
  • Run the relevant test file (pytest tests/.../test_X.py) — not the whole suite, just the affected module.
  • For perf changes (no parity test): write a tiny smoke test that exercises both old/new paths and asserts numerical match against a torch.nn.functional reference.
  • For dtype changes: think about the load path too. Tensor.copy_ silently casts — does that propagate precision loss anywhere? Add a placement-load warning if surprising.
7. Reply on the PR (only if user asks)

If the user wants to push changes back as PR review replies:

bash
gh api repos/OWNER/REPO/pulls/comments/COMMENT_ID/replies \
  -f body="Addressed in commit SHA — explained reasoning in the diff."

Don't auto-reply. The user may want to sanity-check before publishing.

Anti-patterns

  • Pasting a bot's suggestion: block verbatim without checking whether the project already has a different pattern for that situation.
  • "All five comments are valid, fixing all" — at least one is usually wrong or stale. If you don't find a wrong one, you didn't read carefully.
  • Treating "medium priority" as "must-fix" — the bot's priority labels are heuristic. A medium-priority race condition is more important than a high-priority style nit.
  • Editing without reading the test file first — tests document the intended behavior; an edit that breaks the test is usually wrong even if it matches the bot's suggestion.
  • Silent dtype/shape changes — if your fix changes the dtype or shape of a parameter, the load path (apply_placements / state_dict loader) needs to handle the cast and ideally warn on mismatch.

Quick reference: gh CLI cookbook

bash
# List PRs needing attention
gh pr list -R OWNER/REPO --search "review:required"

# Full PR snapshot (metadata + body, no comments)
gh pr view N -R OWNER/REPO --json number,title,state,body,author,headRefName

# Diff for the PR
gh pr diff N -R OWNER/REPO

# Get the commit on the PR head
gh pr view N -R OWNER/REPO --json headRefOid -q .headRefOid

# Resolve / mark conversation as resolved (REST not GraphQL — needs separate call)
# (gh CLI doesn't support this directly; use the GraphQL endpoint or the web UI)

© mingti-org, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/phyai-solve-pr-comments of mingti-org/phyai.

Open the folder on GitHubat commit 36a46bf

Compare with similar skills

Phyai Solve PR Comments next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Phyai Solve PR Comments compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Phyai Solve PR Comments this skillmingti-org/phyai129—~1.9kAutomated safety check: PassMIT
PR Babysitteropeninterpreter/openinterpreter69k3 repos~4.2kAutomated safety check: PassApache-2.0
Diagnosing Superpowers Sessionsobra/superpowers296k3 repos~1.7kAutomated safety check: PassMIT
GitHub Deep Researchbytedance/deer-flow83k5 repos~1.3kAutomated safety check: PassMIT
Greplooponyx-dot-app/onyx32k4 repos~3.3kAutomated safety check: PassMIT
Update V8 Versionopeninterpreter/openinterpreter69k2 repos~845Automated safety check: PassApache-2.0

Similar skills

  • PR Babysitter

    openinterpreter/openinterpreter

    Watches an open GitHub pull request until it merges, handling review comments, diagnosing CI failures and retrying flaky checks along the way.

    69k GitHub starsUsed in 3 repos~4.2k tokens
    DevelopmentAuto-check passed

  • Diagnosing Superpowers Sessions

    obra/superpowers

    Investigates a session where Superpowers went wrong, reads the transcripts on disk and produces an evidence-cited report, optionally prepared as a bug report for the maintainers.

    296k GitHub starsUsed in 3 repos~1.7k tokens
    Agent WorkflowsAuto-check passed

  • GitHub Deep Research

    bytedance/deer-flow

    Researches a GitHub repository over four rounds using the GitHub API and web search, then writes a structured markdown report with timeline, metrics and Mermaid diagrams.

    83k GitHub starsUsed in 5 repos~1.3k tokens
    Research & ScienceAuto-check passed

  • Greploop

    onyx-dot-app/onyx

    Iteratively improves a PR (GitHub), MR (GitLab), or shelved changelist (Perforce) until Greptile gives it a 5/5 confidence score with zero unresolved comments.

    32k GitHub starsUsed in 4 repos~3.3k tokens
    DevelopmentAuto-check passed

  • Update V8 Version

    openinterpreter/openinterpreter

    Bumps the pinned v8 and rusty_v8 versions in Codex, validates the release-candidate path with the v8-canary check, and traces failures to upstream build changes.

    69k GitHub starsUsed in 2 repos~845 tokens
    DevOps & CloudAuto-check passed

  • Last30days

    mvanhorn/last30days-skill

    Research what people actually say about any topic in the last 30 days.

    64k GitHub stars~7.8k tokensUpdated today
    Research & ScienceAuto-check: notes

More from mingti-org/phyai

  • Phyai Local Env Report

    mingti-org/phyai

    Generate a local phyai environment report for debugging system, Python, CUDA/GPU, dependency, workspace package, git, and PHYAI configuration issues.

    129 GitHub stars~590 tokensUpdated 2 days ago
    Auto-check passed

  • Phyai Model Arch Research

    mingti-org/phyai

    A skill your agent uses when the user provides a paper, arXiv link, technical report, model card, checkpoint name, GitHub repository, or local codebase and asks to research, explain, compare, or…

    129 GitHub stars~1.8k tokensUpdated 2 days ago
    Auto-check passed

  • Phyai Model Implement

    mingti-org/phyai

    A skill your agent uses when implementing, porting, integrating, reproducing, or debugging support for a model in PHYAI.

    129 GitHub stars~3.7k tokensUpdated 2 days ago
    Auto-check passed

  • Phyai Communicate With Memory

    mingti-org/phyai

    Analyze a phyai .memory file or directory supplied by the user: parse what task/session it records, locate and inspect the referenced code repository when available, verify claims against local code…

    129 GitHub stars~1.9k tokensUpdated 2 days ago
    Auto-check passed

Works with

Questions about Phyai Solve PR Comments

What does Phyai Solve PR Comments do?

Triage and resolve review comments on a GitHub PR — fetch all comment surfaces (issue / inline / review), validate each suggestion against upstream source rather than trusting blindly, present…. Phyai Solve PR Comments is an agent skill from mingti-org/phyai. Triage and resolve review comments on a GitHub PR — fetch all comment surfaces (issue / inline / review), validate each suggestion against upstream source rather than trusting blindly, present findings to the user before editing, then make focused changes and re-run tests.

When should I use Phyai Solve PR Comments?

Phyai Solve PR Comments fits situations like: the user asks to check PR Ns comments; address review feedback; resolve PR comments.

How do I install Phyai Solve PR Comments in Claude Code?

Run `npx skills add mingti-org/phyai --skill phyai-solve-pr-comments -a claude-code`. Or copy the skill folder (.claude/skills/phyai-solve-pr-comments in mingti-org/phyai) into .claude/skills/phyai-solve-pr-comments in your project. Claude Code loads it when a task matches its description.

How do I install Phyai Solve PR Comments in Codex?

Run `npx skills add mingti-org/phyai --skill phyai-solve-pr-comments -a codex`. Or copy the skill folder (.claude/skills/phyai-solve-pr-comments in mingti-org/phyai) into .agents/skills/phyai-solve-pr-comments in your project. Codex loads it when a task matches its description.

Can I use Phyai Solve PR Comments in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add mingti-org/phyai --skill phyai-solve-pr-comments -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/phyai-solve-pr-comments, .gemini/skills/phyai-solve-pr-comments, .github/skills/phyai-solve-pr-comments and .opencode/skills/phyai-solve-pr-comments in your project.

What does Phyai Solve PR Comments need to run?

Going by SKILL.md and its folder, Phyai Solve PR Comments needs the command-line tools its instructions call (gh, git and pytest).

Does Phyai Solve PR Comments access the network?

SKILL.md contains no URLs. Its commands use gh and git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Phyai Solve PR Comments safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Phyai Solve PR Comments use?

Phyai Solve PR Comments is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Phyai Solve PR Comments use?

About 1.9k tokens (SKILL.md is roughly 7.6k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Phyai Solve PR Comments?

Skills that share tags, products or a category with Phyai Solve PR Comments: PR Babysitter (openinterpreter/openinterpreter, 69k stars), Diagnosing Superpowers Sessions (obra/superpowers, 296k stars), GitHub Deep Research (bytedance/deer-flow, 83k stars) and Greploop (onyx-dot-app/onyx, 32k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Phyai Solve PR Comments?

mingti-org (a GitHub organization) maintains it in mingti-org/phyai, which has 129 GitHub stars. The repository holds 5 skills in this directory. The repository was last updated on October 6, 2026.

Source: mingti-org/phyai on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.