dariomonopoli-dev/Flash 3.5 Action

Run Gemini 3.5 Flash on your repo. Skills-aware, cost-capped, AGENTS.md compatible.

Trust Signals

name	description	required	default
api-key	Gemini API key. Get one at https://aistudio.google.com/apikey.	yes	—
prompt	Prompt for the agent.	yes	—
context	How to assemble repo context. One of: pr, issue, files, repo, none, auto.	no	auto
output	Where to deliver the response. One of: comment, summary, check, file, none. (Pair file with actions/upload-artifact to upload.)	no	comment
output-file	Path for output when output=file or output=artifact.	no	flash-output.md
model	Gemini model. Default is gemini-3.5-flash.	no	gemini-3.5-flash
max-tokens	Hard cap on total input+output tokens for the run. Aborts if exceeded.	no	8000
thinking-budget	Thinking-token cap for thinking-enabled Gemini models (e.g. gemini-2.5+/3.x). 0 disables thinking (most predictable cost). -1 lets the model decide. Positive integer caps thinking at that many tokens. Leave unset for the model default.	no	""
skills	Directory containing SKILL.md files. Auto-loaded into the system prompt.	no	.agent/skills
agents-md	Path to AGENTS.md. Prepended to the system prompt when present.	no	AGENTS.md
pr-number	PR number. Defaults to the event payload.	no	${{ github.event.pull_request.number }}
issue-number	Issue number. Defaults to the event payload.	no	${{ github.event.issue.number }}
github-token	Token for posting comments, creating checks, applying labels.	no	${{ github.token }}

name	description
summary	Full text response from the agent.
error	Error string when the run fails.
artifact-path	Path to the file written when output=file. Pair with actions/upload-artifact to upload.
tokens-used	Total tokens consumed (input + output).
budget-remaining	Tokens left under the max-tokens cap.