Guide · Agent skill
How to use the Bug Report Writer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
The Bug Report Writer is an agent skill from the Claude Code Game Studios setup, published under the MIT licence. It turns a rough description of a game bug into a structured report, or scans a source file for likely defects. This guide explains what a skill is, how to install this one, how to trigger it, and what it needs before you rely on its output.
Bug Report WriterDonchitos/Claude-Code-Game-Studios · MIT Download the skill Quick start
- 1Download and unzip. Download the ZIP from this page and unzip it so you have a bug-report folder containing SKILL.md.
- 2Place the folder. Move the bug-report folder into ~/.claude/skills/ for all projects, or into .claude/skills/ inside one project.
- 3Confirm project files. Make sure .claude/hooks/yaml-helper.sh and .claude/docs/automation-modes.md exist in your project.
- 4Optionally set a test command. Add commands.test to project.yaml if you want verify mode to run related tests.
- 5Run a description. Ask the agent to write a bug report from a plain description of what broke and how to reproduce it.
- 6Review the draft. Open the draft in production/qa/bugs/, check the severity, priority and steps, then submit it yourself.
What an agent skill is and what this one does
An agent skill is a folder of instructions and optional scripts that an AI coding agent loads when a task matches it. This skill is named bug-report and its main file is SKILL.md.
Once installed, the agent can take a description of a game bug and produce a report with a severity, a priority, reproduction steps, expected and actual results, and likely affected files found by searching the codebase. It can also analyze a source file for likely bugs such as null references, off-by-one errors, race conditions and resource leaks, and later check whether a reported fix resolved the bug.
- Description mode: builds a full bug report from your words
- Analyze mode: reads a file and drafts reports for potential bugs
- Verify mode: checks whether a fix removed the problem
- Close mode: closes a bug only after it is marked Verified Fixed
Install the bug-report skill
Download the ZIP from this page and unzip it. Place the bug-report folder in ~/.claude/skills/ to make it available across your projects, or in a project's .claude/skills/ folder to limit it to that project.
If you use the Claude apps instead of Claude Code, upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/
- Project install: .claude/skills/ inside your project
- Claude apps: Settings → Capabilities → Skills
Check the project files the skill needs
The skill expects the Claude Code Game Studios folder layout. It calls .claude/hooks/yaml-helper.sh to read settings, and it reads .claude/docs/automation-modes.md to decide when to ask you questions. Without these files, lookups will fail and results will be generic.
An optional project.yaml can set commands.test so that verify mode runs related tests. If the key is not set, verify mode skips the test run and asks you to confirm manually.
- .claude/hooks/yaml-helper.sh must be present
- .claude/docs/automation-modes.md must be present
- project.yaml is optional; commands.test enables test runs in verify mode
- Bash access is required for the yaml helper and test runs
Trigger the skill with natural requests and commands
You can invoke the skill with the slash command and its arguments, or describe the bug in plain language and let the agent pick the matching mode. The argument hint in the skill lists the four forms.
Give as much detail as you have: what broke, when it happens, how to reproduce it, and what should happen instead. The skill parses this into the report fields.
- Describe a bug: /bug-report followed by your description
- Analyze a file: /bug-report analyze [path-to-file]
- Verify a fix: /bug-report verify [BUG-ID]
- Close a verified bug: /bug-report close [BUG-ID]
/bug-report verify [BUG-ID]Understand the output and where reports are saved
Each report uses a fixed template with a Summary, Classification, Environment, Reproduction Steps, Technical Context, Evidence, Related Issues and Notes. Severity runs from S1-Critical to S4-Low, and Priority runs from P1-Fix this sprint to P4-Won't fix. These labels match the ones used by the bug-triage skill.
Reports are written as drafts to production/qa/bugs/. The skill never files, opens or comments on an issue in any tracker. If the cause is in an engine, a plugin or another vendor's code, the skill adds a draft for that tracker under Notes and stops, so you submit it yourself.
- Reports are saved in production/qa/bugs/
- Severity: S1-Critical, S2-High, S3-Medium, S4-Low
- Priority: P1-Fix this sprint, P2-Fix soon, P3-Backlog, P4-Won't fix
- Analyze-mode results are reasoned from reading code and may include false positives
Safety notes: scripts and network
The skill runs scripts. It calls the yaml helper to read settings, and verify mode can run your test command through Bash. Review the test command in project.yaml before you run verify mode on a project you do not fully trust.
The skill does not use the network, according to its listing. Its drafts are local files in your project, and submission is left to you.
- Runs scripts: yes, including the yaml helper and test runs
- Uses network: no
- Writes draft files only; no automatic submission to trackers
- Automation mode is read from project.local.yaml or project.yaml and defaults to collaborative, which asks before every file write
What people use it for
- Turning a playtester's rough note into a report with numbered reproduction steps and expected versus actual results
- Scanning a combat or inventory script for null references, race conditions or unhandled edge cases before a build
- Checking whether a hotfix removed the code path behind a bug, and recording the result
- Keeping severity and priority labels consistent across a small QA team that uses the bug-triage skill
Frequently asked questions
How do I install the bug-report skill?
Download the ZIP from this page and unzip it. Put the bug-report folder in ~/.claude/skills/ for all projects, or in your project's .claude/skills/ folder for one project. In the Claude apps, you can upload it under Settings → Capabilities → Skills.
Is the bug-report skill free to use?
The skill is published under the MIT licence, as listed on this page. Check the licence file in the source repository for the full terms before you use or redistribute it.
What does the bug-report skill need to run?
It needs Claude Code or a compatible agent with the skill folder installed. The project must include .claude/hooks/yaml-helper.sh and .claude/docs/automation-modes.md, and Bash access is required. A project.yaml with commands.test is optional.
Will it work outside the Claude Code Game Studios folder layout?
Not well. The skill expects paths such as production/qa/bugs/ and a project.yaml config, so projects outside that structure may get generic results or failed file lookups.
Does the skill submit bugs to GitHub or other trackers?
No. It writes drafts only and never files, opens or comments on issues or pull requests. You submit the report yourself.
Can I trust the bugs found in analyze mode?
Treat them as leads, not confirmed bugs. Analyze mode reasons from reading code and can report false positives, so a person should check each one before acting on it.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
This skill suits game developers and QA leads on the Claude Code Game Studios layout who want consistent bug reports with severity and priority labels. It is less useful for projects with different folder structures, for teams that want automatic submission to a tracker, or for treating analyze-mode findings as confirmed defects.
Download the skill