- Home
- Skills
- Agents & Orchestration
- gemini research
gemini research
Offload deep technical research and API lookups to a background Gemini Pro pool with built-in data masking.
Free
Works with the AI Agents you already use
gemini research
Example session with this skill installed
Research the latest rate limits and authentication headers for the Stripe Sigma API. Mask the project name 'Project-Neon'. Provide a structured JSON output for the limits.
- Read your context and instructions
- Compiled the gemini research
I have initiated a background research job for the Stripe Sigma API documentation. I applied a mask for 'Project-Neon' and requested structured output for rate limits. The job has completed and the results are summarized below.
Stripe Sigma API Research Findings
| Category | Technical Detail | Source / Status |
|---|---|---|
| Auth Headers | Stripe-Account, Authorization: Bearer | Official Docs |
| Rate Limit (Standard) | 100 requests per second | Verified |
| Rate Limit (Search) | 20 requests per second | Verified |
| File Uploads | 500 requests per minute | Verified |
| Project Context | Findings isolated from [[PROJECT_1]] | Masked |
Job Metadata
- Job ID:
agy-8829-xql - Conversation ID:
conv-stripe-sigma-v1 - Masking: Active (1 term)
Next steps
- Call
research_report({ from: "agy-8829-xql", path: "docs/stripe-research.md" })to generate a full Markdown report. - If you need specific code examples for these headers, run
start_researchwithresume: "latest". - Check
.research-mask.jsonto add any additional internal service names before further lookups.
Connects securely to your tools. The creator never sees your data.
About this skill
External research tasks often consume your primary model's context window and rate limits. This skill offloads deep documentation dives, API lookups, and technical research to a background Gemini Pro instance, preserving your main agent's quota.
What it does
- Background research executes long-running data gathering tasks without blocking your active session.
- Sensitive data masking replaces organizational names or project titles with deterministic placeholders before they leave your environment.
- Structured data extraction forces research findings into specific JSON schemas for automated parsing of citations or technical specs.
- Continuous conversations allow for iterative follow-ups on thin or dated research by resuming existing conversation IDs.
- Automated reporting generates formatted Markdown reports grounded in research logs directly to your filesystem.
How it works
- Configure masking by placing a
.research-mask.jsonin the repo root to protect sensitive strings. - Initialize research using
start_researchwith a specific prompt, optional schema, and a wait duration. - Monitor progress by polling for results or awaiting the background job completion.
- Iterate findings by resuming the conversation ID if initial data requires more depth or clarification.
- Export results via
research_reportto generate a final grounded deliverable.
Frameworks & tools
This skill utilizes the Gemini/Antigravity MCP server architecture and Google AI Pro models. It manages local state via .research-mask.json and job IDs for persistence.
Why this beats prompting it yourself
Manual research inside a chat session clutters your history and drains your primary API credits. This skill uses a dedicated quota pool and provides a local masking layer that prevents sensitive internal names from being sent to external LLMs.
Use cases
- Researching undocumented API behaviors or rate limits for third-party integrations.
- Gathering competitive technical specs for a product comparison report.
- Performing deep dives into external library documentation without pasting long text blocks.
- Generating structured technical citations for architectural decision records.
Known limitations
Research quality may decrease if the primary subject is masked. It requires a configured Gemini/Antigravity MCP server and cannot access files located within the local repository.
Masking quality varies with abbreviations or context clues. Does not search the local repository. Requires Gemini/Antigravity MCP server access.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 26 days ago
- Free with an Agensi account
Requires