Agent skill

Fix PRs

by fastrepl in fastrepl/anarlog

Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace.

MITAuto-check passedDevelopment

Install Fix PRs

skills CLI
$ npx skills add fastrepl/anarlog --skill fix-prs -a claude-code

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

GitHub CLI
$ gh skill install fastrepl/anarlog fix-prs --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/fastrepl/anarlog.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.agents/skills/fix-prs .claude/skills/fix-prs && 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
fix-prs
GitHub stars
9.5k
Token cost
~2.8k tokens
SKILL.md length
1,475 words
Files
1
Skills in repo
32
Repo updated
First seen
Licence
MIT

At a glance

Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace.

  • Works in 5 steps: Inspect applied stacks, dirty files and… → Read the coordinator's pending merge… → Resolve branches from base toward tip… → …
  • Tasks that involve Failing and flaky tests
  • SKILL.md covers Applicability, Resolve workspace conflicts, Inventory all active PRs and Repair from base upward, plus 1 more section
  • Calls git

What it does

Fix PRs is an agent skill from fastrepl/anarlog. Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace. Resolve conflicts in the current GitButler workspace and fix actionable code review findings and CI failures on its active PRs, including drafts. Validate and publish repairs on their owning branches, resolve addressed review threads, and leave PRs open.

Its SKILL.md is about 2.8k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Development, covering Failing and flaky tests. The repository describes itself as: Open source Granola AI Alternative. The licence is MIT.

When your agent uses it

  • Tasks that involve Failing and flaky tests

Example prompts

  • “/fix-prs”

Workflow steps

5 steps, taken from the first numbered list in SKILL.md.

  1. Inspect applied stacks, dirty files and conflicted commits with
  2. Read the coordinator's pending merge operations, affected branch names and
  3. Resolve branches from base toward tip and each branch's commits oldest-first.
  4. For uncommitted file conflicts, edit the intended content and then mark the
  5. Continue until scoped conflicts are cleared. Make routine resolutions

What it can do on your machine

Read from SKILL.md and the folder at commit 385e117. 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:

    • git

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

  • Network

    No URLs in SKILL.md. Its commands use 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

Fix PRs loads about 2.8k tokens when it runs. Until then it costs about 90 tokens; SKILL.md has 1,475 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~90
When it runs · the whole SKILL.md, loaded when a task matches
~2.8k

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 fastrepl/anarlog at commit 385e117, republished under its MIT licence (© fastrepl). 1,475 words, ~2,761 tokens.

Download SKILL.mdSave it as .claude/skills/fix-prs/SKILL.md (or your agent's skills folder).
name
fix-prs
description
Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace. Resolve conflicts in the current GitButler workspace and fix actionable code review findings and CI failures on its active PRs, including drafts. Validate and publish repairs on their owning branches, resolve addressed review threads, and leave PRs open.

Fix workspace conflicts, PR reviews and CI

Applicability

Use only in the Anarlog repository or its checkouts/worktrees when the active checkout branch is exactly gitbutler/workspace. Before running this workflow, check with git symbolic-ref --quiet --short HEAD (a read-only branch check). On any other branch or detached HEAD, skip this skill and use the normal Git workflow. Do not initialize GitButler or switch branches to activate the skill. Use this checkout's skills, not a global copy or another checkout's skills.

Use this skill when asked to repair workspace conflicts, PR review findings, or CI failures in the Anarlog GitButler repository.

Read the repository-local but skill and the active repository's AGENTS.md, workflow, and testing skills when present. Use but for version control and the repository's preferred authenticated forge tools for PR operations. Derive the repository and configured target from the current workspace.

Invoking this workflow authorizes workspace updates, conflict resolutions and their necessary history rewrites, in-scope fixes, local commits, protected pushes, necessary CI reruns, and evidence-backed review replies/thread resolutions. Creating the skill alone does not execute it. Do not merge, enable auto-merge, create missing PRs, change draft state, or bypass branch protection. Other pushed/shared history rewrites need existing authorization; a new focused commit on the owning branch can avoid that dependency.

Resolve workspace conflicts

  1. Inspect applied stacks, dirty files and conflicted commits with but status --json (use -fv when file details are needed). Conflict repair covers all applied workspace branches by default, including branches without PRs; honor a narrower user scope. Record all local branch heads before any updates or repairs so rewritten descendants can be identified afterward. Coordinate with active agents before shared checkout mutations and preserve their unrelated edits.
  2. Read the coordinator's pending merge operations, affected branch names and reviewed heads before updating. While any pending merge stack remains applied, skip the workspace-wide but pull; but pull --check does not make that pull selective. Exclude pending stacks from repairs, and run independent repairs only when every GitButler mutation and ancestor push scope can demonstrably leave those stacks untouched. Otherwise defer repairs until their merge operations reach a confirmed terminal state. Once this guard is satisfied, refresh from the configured target with but pull. When other agents' branches are applied or scope is narrower than the workspace, first run but pull --check to identify affected branches. The default workspace-wide invocation authorizes routine in-scope updates and conflict repairs, subject to AGENTS.md's approval boundaries. Ask before resolving semantic conflicts, dependency updates, generated-file conflicts, or conflicts involving another person's work unless existing user authorization explicitly covers that particular resolution. Also ask before modifying work outside the authorized scope. Keep the target and stack dependencies unchanged. If dirty changes block an update, use the but skill's parking/restoration procedure only for in-scope work.
  3. Resolve branches from base toward tip and each branch's commits oldest-first. Use the installed but skill's conflict tools, preferably but resolve conflicts <full-branch-name> and but resolve apply <path>:<conflict-number> --commit <full-branch-name> with merged content on stdin or --file. Read base, ours and theirs and preserve both intended behaviors. Resolution rewrites commit IDs; use branch names and fresh output. Do not amend conflicted commits or use raw Git writes.
  4. For uncommitted file conflicts, edit the intended content and then mark the paths resolved with but resolve <path>.... Keep the resulting changes on their owning branch when ownership is established; preserve ambiguous dirty work rather than commit it into a guessed branch. Regenerate generated files through their owning tasks when required instead of hand-editing artifacts.
  5. Continue until scoped conflicts are cleared. Make routine resolutions directly; if the two sides require an unresolved product or architectural decision, report the exact alternatives and continue independent repairs. Run relevant checks on the resolved code and refresh the branch/PR map and heads before diagnosing reviews or CI. Track remaining conflicts as blockers even when every remote PR check is green.

Publish resolved branches that already have active PRs with the consolidated repair push below. Keep branches without PRs local unless separately authorized to publish them, and inspect ancestor push scope so they are not sent implicitly.

Inventory all active PRs

  • Map applied stacks with but status --json and reconcile exact head branches against live open PRs, including drafts. Scope is workspace PRs, not every open PR in the repository. Record base-to-tip dependencies, PR URLs and remote SHAs.
  • For every PR, paginate submitted reviews, review threads and their comments, conversation comments, check runs and commit statuses. Include unresolved outdated findings, review summaries and linked reviewer findings. Read failed job logs and inspect pending/cancelled checks; green CI alone says nothing about unresolved reviews. Inaccessible sources are coverage gaps.
  • Keep a findings ledger keyed by PR plus stable thread/comment/check ID: source, issue, owning branch/commit, verified head, disposition, evidence, next action. Deduplicate reports while retaining their links. Reconcile all outstanding findings on each round, including previously resolved findings whose fix is still unverified. Do not limit later passes to new comments.
  • Trace each finding at its PR's published head and through the entire stack, including upper PRs. A fix higher in the stack is a dependency, not proof the earlier PR is fixed. Verify that fix and its tests, record its PR/SHA, and do not duplicate it or resolve the lower thread as fixed on its own head.
Show full SKILL.md (596 more words)Show less

Repair from base upward

  1. Work in batches of about 10 PRs, nearest trunk first, with full-stack context. Complete independent repairs while recording blocked PRs. Validate reviewer claims against actual behavior; do not blindly implement every suggestion. Dismiss a false positive only with specific code/test evidence. Escalate an unresolved product decision while continuing unrelated fixes.
  2. Classify CI failures from logs. Fix PR-caused failures on the branch that introduced them. Compare against target-branch failures before changing code for an inherited break. Keep infrastructure, secret/permission, and external service failures explicit; do not weaken checks or broaden credentials. Rerun an unchanged failure once only when there is evidence of a transient failure or a repaired external prerequisite. Repeated identical failures need diagnosis or a blocker, not a rerun loop.
  3. Preserve unrelated dirty files and other agents' commits. Inspect fresh but diff IDs and make the smallest complete change on the existing owning branch, with meaningful regression coverage and repository-required checks. Amend an appropriate unpublished commit; otherwise create a focused commit on that branch. Never invent a repair branch stacked above the issue.
  4. Compare all local branch heads with the pre-repair snapshot, including descendants rewritten by updates or repairs without direct file edits. For each affected stack, select its topmost affected branch; pushing it publishes that branch and its ancestors, not its descendants. Check every included branch's destination and current remote head, then inspect scope with but push <topmost-affected-branch> --dry-run. Obtain authorization for any otherwise unrelated publication or history rewrite before including it. Consolidate fixes and push once per affected stack through that branch. Verify every affected PR's published head, including rewritten descendants. Do not overwrite newer remote work, bypass hooks, or skip force-push protections. Refresh IDs after mutations when the next operation needs them.

Repeat conflict repair after updates or mutations introduce new conflicts. Being behind trunk alone is not a reason to publish equivalent rewritten history.

Verify and finish

  • Re-read every affected remote head after pushing and verify each fix at the published SHA with relevant checks. Attach PRs being updated with mcp__codex_app__attach_artifact when available.
  • Resolving addressed review threads is a required part of fixing PRs. For each verified fix, reply briefly with the change and commit/test evidence, then resolve the thread through the forge. Explain evidence-backed false positives before resolving them too. Include human and automated reviews, including outdated unresolved threads. A push, passing check, reply, or outdated flag does not resolve a thread. Re-fetch threads to confirm resolution was saved. Leave unverified or blocked findings open and report their links and next action. If resolution fails or permissions are missing, report that blocker and carry the pending resolution into the follow-up.
  • Check live CI, reviews and the complete ledger again for all scoped PRs. New heads invalidate old check conclusions. Repairs, local tests, published fixes, remote CI and resolved reviews are separate facts. Completion requires no unresolved workspace conflicts, no actionable findings, no addressed review threads left unresolved, and successful applicable checks on current heads; pending checks, inaccessible findings, dependencies and external failures remain explicit.
  • Finish available repairs before an external wait. Follow the but skill's continuation procedure: reuse/create a thread heartbeat for pending CI/review verification, save repo, PRs, heads, ledger, conflict state, completed repairs and blockers, and recheck live state on resumption. Notify only for meaningful changes; pause once complete or waiting only for user input. Do not watch, sleep or repeatedly poll. Never claim a follow-up exists if scheduling failed.

Report conflict resolutions, fixes and commits, local validation, published heads, review resolutions, and remaining conflicts/CI/findings with links. Leave all PRs open.

© fastrepl, 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 .agents/skills/fix-prs of fastrepl/anarlog.

Open the folder on GitHubat commit 385e117

Compare with similar skills

Fix PRs 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.

Fix PRs compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Fix PRs this skillfastrepl/anarlog9.5k—~2.8kAutomated safety check: PassMIT
PR Babysitteropeninterpreter/openinterpreter69k3 repos~4.2kAutomated safety check: PassApache-2.0
Systematic Debuggingultralisp/ultralisp25851 repos~2.4kAutomated safety check: PassNone
Iterate PRmeshery/meshery-operator1518 repos~2.2kAutomated safety check: PassApache-2.0
Exposed Bug Fix WorkflowJetBrains/Exposed9.3k—~3.8kAutomated safety check: PassApache-2.0
React Router Bug Fix Workflowremix-run/react-router57k—~1.3kAutomated safety check: PassMIT

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
  • Systematic Debugging

    ultralisp/ultralisp

    A skill your agent uses when encountering any bug, test failure, or unexpected behavior, before proposing fixes

    258 GitHub starsUsed in 51 repos~2.4k tokens
    DevelopmentAuto-check passed
  • Iterate PR

    meshery/meshery-operator

    Iterate on a PR until CI passes. An agent skill from meshery/meshery-operator.

    151 GitHub starsUsed in 8 repos~2.2k tokens
    DevelopmentAuto-check passed
  • Exposed Bug Fix Workflow

    JetBrains/Exposed

    Official

    Takes a GitHub or YouTrack issue for the Exposed project through reproduction, a failing test, a fix, validation and a pull request.

    9.3k GitHub stars~3.8k tokensUpdated yesterday
    DevelopmentAuto-check passed
  • React Router Bug Fix Workflow

    remix-run/react-router

    Fixes a React Router bug reported in a GitHub issue end to end: fetching the issue, validating the reproduction, writing a failing test and implementing the fix on a new branch.

    57k GitHub stars~1.3k tokensUpdated 2 days ago
    DevelopmentAuto-check passed
  • Systematic Debugging

    ChrisWiles/claude-code-showcase

    Applies a four-phase debugging routine that finds the root cause of a bug or failing test before any fix is written.

    6.1k GitHub starsUsed in 3 repos~1.2k tokens
    DevelopmentAuto-check passed

More from fastrepl/anarlog

All 32 skills in this repo
  • But

    fastrepl/anarlog

    Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace.

    9.5k GitHub stars~5.8k tokensUpdated yesterday
    Auto-check passed
  • Create PRs

    fastrepl/anarlog

    Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace.

    9.5k GitHub stars~1.2k tokensUpdated yesterday
    Auto-check passed
  • Fix Ready PRs

    fastrepl/anarlog

    Inspect every open non-draft PR for CI failures and unresolved Cursor Bugbot findings, then fix them on the existing PR branches.

    9.5k GitHub stars~1.4k tokensUpdated yesterday
    Auto-check passed
  • No Use Effect

    fastrepl/anarlog

    Avoid direct React useEffect usage when writing or reviewing React components and hooks.

    9.5k GitHub stars~825 tokensUpdated yesterday
    Auto-check passed
  • QA Critical UX

    fastrepl/anarlog

    QA Anarlog's critical Pro user journey on a signed staging candidate — onboarding, responsive launch, microphone and system-audio capture, automated summaries, and cloud sync.

    9.5k GitHub stars~1.4k tokensUpdated yesterday
    Auto-check passed
  • Reactive Sqlite UI

    fastrepl/anarlog

    Build SQLite-backed reactive UI in apps/desktop using stable patterns for reads, selection, forms, writes, and loading states.

    9.5k GitHub stars~699 tokensUpdated yesterday
    Auto-check passed

Categories

Questions about Fix PRs

What does Fix PRs do?

Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace. Fix PRs is an agent skill from fastrepl/anarlog. Use only in the Anarlog repository when the active checkout branch is exactly gitbutler/workspace.

When should I use Fix PRs?

Fix PRs fits situations like: tasks that involve Failing and flaky tests.

How do I install Fix PRs in Claude Code?

Run `npx skills add fastrepl/anarlog --skill fix-prs -a claude-code`. Or copy the skill folder (.agents/skills/fix-prs in fastrepl/anarlog) into .claude/skills/fix-prs in your project. Claude Code loads it when a task matches its description.

How do I install Fix PRs in Codex?

Run `npx skills add fastrepl/anarlog --skill fix-prs -a codex`. Or copy the skill folder (.agents/skills/fix-prs in fastrepl/anarlog) into .agents/skills/fix-prs in your project. Codex loads it when a task matches its description.

Can I use Fix PRs 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 fastrepl/anarlog --skill fix-prs -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/fix-prs, .gemini/skills/fix-prs, .github/skills/fix-prs and .opencode/skills/fix-prs in your project.

What does Fix PRs need to run?

Going by SKILL.md and its folder, Fix PRs needs the command-line tools its instructions call (git).

Does Fix PRs access the network?

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

Is Fix PRs 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 Fix PRs use?

Fix PRs 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 Fix PRs use?

About 2.8k tokens (SKILL.md is roughly 11k 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 Fix PRs?

Skills that share tags, products or a category with Fix PRs: PR Babysitter (openinterpreter/openinterpreter, 69k stars), Systematic Debugging (ultralisp/ultralisp, 258 stars), Iterate PR (meshery/meshery-operator, 151 stars) and Exposed Bug Fix Workflow (JetBrains/Exposed, 9.3k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Fix PRs?

fastrepl (a GitHub organization) maintains it in fastrepl/anarlog, which has 9,452 GitHub stars. The repository holds 32 skills in this directory. The repository was last updated on October 8, 2026.

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