This 42-prompt list turns Gemini 3.6 Flash's strengths into reproducible task contracts: put the complete corpus into its 1M context, require line-by-line citations, define an explicit output schema, and control costs with independent verification and tiered routing.
Suitable tasks: full-document retrieval, contract/policy checking, video and audio analysis, browser automation, codebase audits, structured batch processing, and sourced research briefs.
Unsuitable tasks: tasks that require the model to decide permissions on its own, automatically submit irreversible actions, or make open-ended factual judgments without verifiable inputs.
Applicable model version: Gemini 3.6 Flash; page data is current as of July 2026.
Applicable client, agent, or API: Gemini API, Google AI Studio, and agent harnesses with file/screen tools; the browser prompts assume that the harness can provide screen access.
Recommended reasoning tier and parameters: the source recommends using output contracts and citation constraints; it does not provide a unified set of API parameters. For code or agent tasks, start with thinking_level=medium, move to high for difficult tasks, and record the additional tokens.
The following are representative complete prompts that can be copied directly from the original; the original also contains 42 templates. These were selected to cover the model's main input modalities and agent tool types.
I am attaching [N] complete documents totaling roughly [SIZE]: [ATTACH THE FULL SET].
Question: [WHAT YOU NEED TO KNOW]
Answer from the attached material only. For every claim, cite the document name and the section or page it came from. If the answer requires combining facts from multiple documents, show the chain. If the corpus does not contain the answer, say so plainly rather than reasoning from general knowledge.Read this complete document set: [ATTACH ALL].
Find every place where two documents state incompatible things: different numbers for the same metric, conflicting policies, dates that cannot both be true, procedures that contradict each other.
Output a table: claim A (with source), claim B (with source), type of conflict, and which one is more likely authoritative based on recency or document status. Do not report differences that are just different levels of detail — only genuine contradictions.Watch this video and produce a working digest: [ATTACH VIDEO].
Deliver: a one-paragraph summary, a timestamped outline of the major sections, every claim or number stated on screen or in narration (with timestamp), any on-screen text worth capturing, and the three moments most worth watching directly.
Timestamps must be accurate — I will spot-check them.Complete this task in the browser: [DESCRIBE THE TASK].
Done means: [MEASURABLE END STATE — e.g. the form is submitted and the confirmation number is captured].
Rules: take a screenshot before any irreversible action and describe what you are about to do. Never enter payment details, delete data, or send anything without explicit confirmation. If a page differs from what you expected, stop and describe the difference rather than improvising.
Report the confirmation evidence at the end, not just a claim of success.Here is the complete repository: [ATTACH]. Symptom: [WHAT HAPPENS]. Expected: [WHAT SHOULD HAPPEN].
Trace the actual execution path through the real code — not a plausible one. Identify the exact file and line where behavior diverges from intent, explain why, and give the minimal fix.
If multiple causes are consistent with the symptom, rank them and give the one-line check that distinguishes them. Do not propose rewriting modules that are not implicated.Extract structured data from the following item: [PASTE OR ATTACH ITEM].
Return JSON matching exactly this shape: [PASTE YOUR SCHEMA].
Rules: never invent a value — use null for anything not present in the source. For every extracted field, the value must be traceable to specific text in the input. If the input appears to be a different document type than expected, return {"error": "unexpected_type"} and nothing else.Research this question and build a brief: [THE QUESTION]. Decision it informs: [WHY IT MATTERS].
Structure: what is established (with source type), what is contested (both positions, one line each), what is unknown, and your synthesis with a confidence level.
Every number gets a source. Separate facts from your inference explicitly. Where you are unsure, write "unsure" — do not fill the gap with plausible filler. I would rather have six solid facts than twenty vibes.The three directly adoptable workflows given in the original:
Document intelligence: Whole-Corpus Question → Contradiction Finder → Needle Audit; read everything first, then exhaustively verify contradictions and key terms.
Multimodal to action: Recording-to-Actions → Structured Extractor → Bulk Quality Gate; turn audio/video decisions into task records with a schema, then apply a quality gate.
Batch content: Voice-Matched Draft → Ruthless Edit (new session) → Repurposing Fan-Out; separate generation, editing, and multi-platform rewriting to avoid publishing the first-round output directly.
The original explicitly recommends putting the complete corpus into context rather than pasting only pre-filtered excerpts.
Every extraction template requires source pointers, null, or unsure, making the distinction between “no evidence” and “the model does not know” explicit.
The browser template requires a screenshot before irreversible actions, confirmation before final submission, and a stop when the page deviates from expectations.
The original positions Gemini 3.6 Flash as a volume tier and recommends upgrading the hardest 20% of tasks to a stronger model; this is routing advice, not an official guarantee.
These are reusable prompts, not controlled experimental results from PromptsRush for each task; they should be regression-tested against your own files, permissions, and tools.
Browser prompts cannot replace client-side permission controls; “stop and report” must be enforced by the harness.
A 1M context and video/audio capabilities do not guarantee successful processing for every file type, duration, or size; test against the API's current limits.
Conclusions about cache savings and output tokens depend on provider billing and the article's prices cannot be applied directly to other routes.
The original states the core rule as “Cite or say unsure”; this should be used together with complete sources, field schemas, and human verification.
Gemini 3.6 Flash