ihealthstrategy/ReviewPilot

AI code review for a pull request using Claude Code (agentic, whole-project context).

Trust Signals

name	description	required	default
engine	Review engine: claude-agent (recommended) \| claude-code \| mock	no	claude-agent
anthropic-api-key	Anthropic API key for the Claude engine.	no	—
anthropic-base-url	Override the Anthropic API base URL (for a relay/gateway). Empty = api.anthropic.com.	no	""
anthropic-model	Main model name (set to a model your relay/gateway exposes; empty = client default).	no	""
anthropic-small-fast-model	Cheap model for background tasks (Claude Code uses one); set to a model your relay exposes.	no	""
anthropic-auth-token	Bearer auth token, if the relay expects Authorization: Bearer instead of x-api-key.	no	""
agent-model	Model for the claude-agent engine (empty = SDK default).	no	""
insight-file	Path to a project-understanding cache file (persist with actions/cache).	no	""
fail-on-severity	Fail the check (block merge) when a finding is this severity or worse: info\|minor\|major\|critical. Empty = never fail.	no	""
check-run-name	Name of the published status check.	no	ReviewPilot
only-changed-lines	Keep only findings on lines the PR changed (noise reduction): true\|false.	no	false
output-language	Language for the human-readable finding text (e.g. 'Chinese'). Empty = English.	no	""
engine-args	Override args for the CLI engine (claude-code), e.g. '-p --output-format text --dangerously-skip-permissions'.	no	""
github-token	Token for PR comments + check runs and pulling the action image. Pass the workflow github.token.	no	""
image	Prebuilt action image to run.	no	ghcr.io/ihealthstrategy/review-pilot-action:v1

name	description
summary	Markdown review summary (the PR comment body).
findings	Number of findings.
conclusion	Check conclusion: success \| neutral \| failure.