victorarias/shitty-reviewing-agent

Agent-driven PR reviewer using LLMs

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Stale
License
None
namedescriptionrequireddefault
providerLLM provider (openrouter, gemini, anthropic, openai, etc.). Aliases: gemini→google, vertex→google-vertex.no
api-keyAPI key for the LLM provider (required for non-Vertex providers; optional for Vertex AI)no
modelModel to use (e.g., anthropic/claude-sonnet-4)no
compaction-modelOptional model to use for context compaction summaries (defaults to gemini-3-flash-preview for Google, else same as model)no
fallback-providerFallback LLM provider (defaults to primary provider if only fallback-model is set)no
fallback-modelFallback model to use when the primary model hits sustained rate limitsno
fallback-api-keyAPI key for the fallback provider (defaults to primary api-key if omitted)no
max-filesMax files to review (skips if exceeded)no
ignore-patternsComma-separated glob patterns to skipno
debugEnable debug loggingno
reasoningThinking level (off, minimal, low, medium, high, xhigh)no
temperatureSampling temperature (0-2). Gemini 3 defaults to 1.0 and may behave unexpectedly at lower values.no
allow-pr-toolsAllow PR-creation tools in PR review mode (default false; always allowed in schedule mode).no
experimental-pr-explainerExperimental: post a PR review guide comment and one explainer comment per changed file.no
bot-nameBot/app mention name for @bot command triggers (e.g., my-app).no
app-idGitHub App ID (optional; default uses implicit GITHUB_TOKEN)no
app-installation-idGitHub App installation IDno
app-private-keyGitHub App private key PEMno

no outputs