Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Pr Sweep

ASecurity

Review every open PR with the code-reviewer agent and an adversarial review, apply the findings, and merge each one once CI is green. Use when asked to ship, sweep, or merge open PRs.

22 stars
0 votes
0 copies
2 views
Added 9/23/2026
ai-agentsgojavabashcode-reviewgit

Works with

claude code

Security Analysis

A100/100

Scanned 9/29/2026

$npx -y skills add Chemaclass/agnostic-ai --skill pr-sweep --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Pr Sweep?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Pr Sweep
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/chemaclass-pr-sweep/badge)](https://www.skillsdirectory.com/skills/chemaclass-pr-sweep)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: pr-sweep
description: Review every open PR with the code-reviewer agent and an adversarial review, apply the findings, and merge each one once CI is green. Use when asked to ship, sweep, or merge open PRs.
argument-hint: "[PR-number ...] [--no-merge]"
disable-model-invocation: false
---

# PR sweep

Take every open PR from review to merge without asking between steps. With PR numbers, sweep only those. With `--no-merge`, stop at green PRs.

## 1. List

```bash
gh pr list --state open --json number,title,headRefName,additions,deletions,author
```

Skip PRs by other authors and drafts; report them. Empty list: stop.

## 2. Review

Run two reviews on the PR branch, in parallel:

- The `code-reviewer` agent for Go diffs (other diffs against the project rules). Keep its Opus model; never pass a cheaper override.
- An adversarial review that challenges the design, not only the lines: `/codex:adversarial-review --base origin/main` where the Codex plugin is installed (Claude Code). Without it, run a second reviewer pass told to question the approach, its assumptions, and how it fails in real use.

Read every result before editing. After each round of fixes, run the adversarial review again until it approves or every remaining finding is rejected in step 4.

## 3. Verify each finding

A finding is a claim, not a task. Before acting on one:

- Reopen the file and line it cites on the PR head.
- Reproduce behavior claims with a built binary or a failing test.
- Check vendor claims against the vendor page, not memory.

Keep what holds. Record what does not, with the reason.

## 4. Apply

Check out the PR branch in the main checkout (worktree agents cannot run git while the RTK hook rewrites it) and pull. For each confirmed finding:

- Fix it on the branch. Test first for behavior changes: the new test must fail on the old code.
- A finding that needs a different design or touches other targets gets its own issue (`gh issue create --assignee Chemaclass` with a label). Do not widen the PR.
- A finding you reject after verification is answered, not ignored.

Then post one PR comment listing the commit that applied the findings, the issues filed, and each rejection with its reason.

## 5. Gate

Before every push, all green. Check each command by its own exit code; a pipe such as `make ci-local | grep FAIL | head` exits 0 when the gate fails.

- `make ci-local`; use `SKIP_JETBRAINS=1` only when Java or Gradle is unavailable, report the skip, and check JetBrains CI when its inputs changed
- `make site-test` when site content changes, counting skips, not the exit code
- `./agnostic-ai sync --check`. Drift only in gitignored generated files after a fresh checkout is a stale ledger: run `sync`, confirm 0 tracked files change.
- CHANGELOG bullets pass the length check in `.agnostic-ai/agents/changelog-curator.md`.

Commit with conventional commits, GPG-signed, no AI attribution or session links. Push.

PR checks run Go tests on Linux only. When the PR touches paths, renames, permissions, file watching, or import, dispatch the full matrix on the branch (`gh workflow run ci.yml --ref <branch>`) and wait for it on the head SHA before merging.

## 6. Merge

For each PR, oldest first, or the one that moves shared files (`sources.lock`, `signals.tsv`, `CHANGELOG.md`) first:

1. Wait for checks on the current head SHA: `gh pr checks <N>`. Missing or pending checks are not green.
2. If `mergeStateStatus` is not `CLEAN`, merge `origin/main` into the branch (never rewrite pushed history), keep both sides of shared docs, rerun the gate, push, and wait again.
3. `gh pr merge <N> --squash --admin --delete-branch`. If `--admin` is rejected, use `--auto --squash --delete-branch` and report that the PR awaits approval.
4. Fast-forward local `main` from `origin/main` and delete the local branch. Stop if `main` has unpublished commits; never reset them away.

PR Go tests run on Linux only. After the last merge, wait for the `CI` workflow on the new `main` head and confirm every job passed, Windows included. A red main is the first thing to fix.

## 7. Report

One table: PR, merge commit, findings applied, findings moved to issues, findings rejected. Then the open follow-up issues.

## Stop

Stop and report when a finding cannot be verified either way, CI stays red after one fix, or a merge is blocked even for `--auto`.

Attribution

ChemaclassChemaclass
View sourceSee grades on GitHubMore from Chemaclass →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698431 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →