aigne-io/AIGNE CodeSmith

AI-powered code reviews and PR summaries using Anthropic Claude. Get intelligent, context-aware feedback on your pull requests.

View on GitHub

Trust Signals

Scorecard Score: not yet scored
Maintenance Recency: Maintainedlast commit Nov 5, 2025
License: MIT

Pinned Snippet

workflow.ymlSHA-pinned

uses: aigne-io/aigne-code-smith@07548712d7dcce2341f0ca8274f7868490b834f9 # v0.1.6

tags can be moved; commit SHAs can't. why a SHA?

Inputs

name	description	required	default
debug	Enable debug mode	no	false
max_files	Max files to summarize and review. Less than or equal to 0 means no limit.	no	150
review_simple_changes	Review even when the changes are simple	no	false
review_comment_lgtm	Leave comments even if the patch is LGTM	no	false
path_filters	The path filters, e.g., "src/.py", "!dist/", each line will be considered as one pattern. See also - https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#onpushpull_requestpull_request_targetpathspaths-ignore - https://github.com/isaacs/minimatch	no	!dist/ !/.app !/.bin !*/.bz2 !*/.class !*/.db !*/.csv !*/.tsv !*/.dat !*/.dll !*/.dylib !*/.egg !*/.env !*/.glif !*/.gz !*/.xz !*/.zip !*/.7z !*/.rar !*/.zst !*/.ico !*/.jar !*/.tar !*/.war !*/.lo !*/.log !*/.md !*/.mp3 !*/.wav !*/.wma !*/.mp4 !*/.avi !*/.mkv !*/.wmv !*/.m4a !*/.m4v !*/.3gp !*/.3g2 !*/.rm !*/.mov !*/.flv !*/.iso !*/.swf !*/.flac !*/.nar !*/.o !*/.ogg !*/.otf !*/.p !*/.pdf !*/.doc !*/.docx !*/.xls !*/.xlsx !*/.ppt !*/.pptx !*/.pkl !*/.pickle !*/.pyc !*/.pyd !*/.pyo !*/.pub !*/.pem !*/.rkt !*/.so !*/.ss !*/.eot !*/.exe !*/.pb.go !*/.lock !*/.ttf !*/.yaml !*/.yml !*/.cfg !*/.toml !*/.ini !*/.mod !*/.sum !*/.work !*/.json !*/.mmd !*/.svg !*/.jpeg !*/.jpg !*/.png !*/.gif !*/.bmp !*/.tiff !*/.webm !*/.woff !*/.woff2 !*/.dot !*/.md5sum !*/.wasm !*/.snap !*/.parquet !/gen/ !/_gen/ !/generated/ !/@generated/ !/vendor/ !*/.min.js !*/.min.js.map !*/.min.js.css !*/.tfstate !*/.tfstate.backup !/package.json !/package-lock.json !/yarn.lock !/pnpm-lock.yaml
disable_review	Only provide the summary and skip the code review.	no	false
disable_release_notes	Disable release notes	no	false
anthropic_base_url	The url of the openai api interface.	no	https://api.anthropic.com/v1
anthropic_light_model	Model to use for simple tasks like summarizing diff on a file.	no	claude-sonnet-4-20250514
anthropic_heavy_model	Model to use for complex tasks such as code reviews.	no	claude-sonnet-4-20250514
anthropic_model_temperature	Temperature for GPT model	no	0.05
anthropic_retries	How many times to retry OpenAI API in case of timeouts or errors?	no	5
anthropic_timeout_ms	Timeout for OpenAI API call in millis	no	360000
anthropic_concurrency_limit	How many concurrent API calls to make to OpenAI servers?	no	6
github_concurrency_limit	How many concurrent API calls to make to GitHub?	no	6
system_message	System message to be sent to OpenAI	no	You are `@codesmith` (aka `github-actions[bot]`), a language model trained by Anthropic. Your purpose is to act as a highly experienced software engineer and provide a thorough review of the code hunks and suggest code snippets to improve key areas such as: - Logic - Security - Performance - Data races - Consistency - Error handling - Maintainability - Modularity - Complexity - Optimization - Best practices: DRY, SOLID, KISS CRITICAL REVIEW FOCUS: Only post comments for CRITICAL issues that significantly impact code quality: - Security vulnerabilities or potential exploits - Logic errors that could cause bugs or incorrect behavior - Performance issues that could impact system efficiency - Error handling problems that could cause crashes or data loss - ALL language quality issues are considered CRITICAL: * Spelling errors in comments, strings, and documentation * Grammar mistakes and awkward phrasing * Unclear or confusing wording that may hinder understanding * Variable/function names that could be more descriptive or follow better naming conventions * Comment clarity and proper English usage STRICT FILTERING: Do NOT comment on minor style preferences, formatting choices, or subjective improvements unless they fall into the CRITICAL categories above. Help improve code readability for international teams by treating all language improvements as critical for professional code quality.
summarize	The prompt for final summarization response	no	Provide your final response in markdown with the following content: - Walkthrough: A high-level summary of the overall change instead of specific files within 80 words. - Changes: A markdown table of files and their summaries. Group files with similar changes together into a single row to save space. Avoid additional commentary as this summary will be added as a comment on the GitHub pull request. Use the titles "Walkthrough" and "Changes" and they must be H2.
suggest_pr_title_prompt	The prompt for generating pull request title	no	Based on the following git diff, generate 3 conventional commit style pull request titles: $diff Requirements: - Follow conventional commit format - Be concise and descriptive - MUST use English language only for pull request titles - Use proper English grammar and spelling - Return only a numbered list of titles, nothing else
summarize_release_notes	The prompt for generating release notes in the same chat as summarize stage	no	Craft concise release notes for the pull request. Focus on the purpose and user impact, categorizing changes as "New Feature", "Bug Fix", "Documentation", "Refactor", "Style", "Test", "Chore", or "Revert". Provide a bullet-point list, e.g., "- New Feature: Added search functionality to the UI". Limit your response to 50-100 words and emphasize features visible to the end-user while omitting code-level details.
language	ISO code for the response language	no	en-US
bot_icon	The icon for the bot	no	<img src="https://avatars.githubusercontent.com/u/143965177?s=48&v=4" alt="Image description" width="15" height="15">
suggest_pr_title	Generate conventional style titles for the pull request.	no	false

Outputs

no outputs