A RESEARCH TOOLKIT FOR AI AGENTS
Turn questions or existing material into reports with concrete judgments and traceable sources. Lite and Full share one research baseline.
The central trade-off is how much work remains after sending a command: checking the resulting state, resolving uncertain outcomes and maintaining the integration when the application changes. That work determines whether a successful demonstration becomes a repeatable business process.
Report excerpt
Original English research, with primary sources, a retained first draft, and a documented review correction.
Section 5 asks when a hybrid API-and-browser workflow should change.
The review identified a missing counterexample: a working browser step can remain worthwhile when replacing and revalidating it costs more than the expected savings.
There is no benefit in preserving this split if the API later covers document generation with adequate permissions and result checking.
The draft treated closing a coverage gap as sufficient reason to retire the browser step.
Full passage ↗If the API later covers document generation with adequate permissions and result checking, the original coverage gap disappears. Retiring the browser step then depends on whether the expected maintenance and recovery savings justify migration and revalidation costs.
The revision separates capability coverage from the economics of changing a working system.
Full passage ↗Public-source analysis, 10 October 2026. The workflow is illustrative; no deployment or comparative performance experiment was conducted. This case does not establish a general Toolkit quality advantage.
See an existing coding-agent procurement report, the source details it missed, and the revisions those checks produced. Original drafts and limits remain available.
Explore the report update →RESEARCH REQUEST
Choose Lite where author checks fit the use. Choose Full to bind selected reviews and revision decisions to the delivery version. Let your agent recommend from the task; length is secondary.
If repository access is unavailable, upload the files or paste the instructions. Native Skill installation is optional.
Ways to provide the toolkitComplete workflow (Full) · guide
Use Research Toolkit: https://github.com/rrrrrredy/research-toolkit Read SKILL.md and load the methods by stage. Research question: [Include the question, audience, and intended use.] Scope and length: [Include scope, information cutoff, length and counting unit.] Existing material: [Attach files or links, including any existing report.] Recommend Lite or Full for the task, including whether reviews and revisions need to stay bound to the delivery version; explain why and respect my choice. Use the context to establish the brief; ask only about consequential gaps. Read required sources, check important claims and calculations, and develop comparisons, counterevidence and conclusions for the intended reader. Reuse existing research and correct affected passages when material changes. Lite completes the author checklist; Full adds the chosen review, revision and delivery checks. State the review actually performed. Deliver the complete report with usable citations and a short change note when revising. Maintain working records yourself, separately from the report. Use MCP/CLI for saved progress and checks when connected.
Replace the bracketed text with your research request. Include any known audience, scope, length, time cutoff, or required sources.
Setup for your agentFrom the brief to the report, each stage has a concrete output.
Audience, scope, decisions, and an agreed outline.
Research brief ↗Primary sources, access limits, and contrary evidence.
Source records ↗Link facts to explanations, tradeoffs, and implications.
Claim records ↗Check evidence and reader usefulness in both workflows; Full adds the selected review.
Review records ↗Concrete judgments, traceable sources, and honest limits.
Complete report ↗The links open this case's brief, source and claim records, complete reviews, and final article.
Scope, sources, reasoning, writing, and review. Detailed guidance is loaded by your agent when needed.
Save tasks and progress, run the chosen reviews, and check delivery consistency.
Choose the input your interface supports. Plugin and MCP tools are optional.
| Repository link | Provide the repository URL and research request. The agent needs to read SKILL.md and the required references. | Repository use |
|---|---|---|
| Files or attachments | Provide SKILL.md and the reference files from the downloaded repository. Supply further methods when needed. | Download files |
| Pasted text | Paste the full entry instructions and the reference sections needed for the task. A URL alone is insufficient without repository access. | Chat-only use |
| Native Skill | For repeated use, install the complete toolkit through the host’s supported Skill mechanism. | Skill setup |
| Plugin | A bundle of the methods and local MCP tools for hosts that support its package format and local execution. | Installation guide |
|---|---|---|
| MCP | Eight local tools for task records, progress, guidance, draft checks, review calls, and delivery checks. | Connect MCP |
Reading or uploading the methods does not start MCP or run an independent review. The tools need local stdio MCP support; the plugin also requires its package format.
Full uses a degraded self-review when no reviewer backend is configured. Choose an external command or compatible endpoint for another opinion; Astra in a separate context can provide independent review, including for an author using the same model; no additional outside reviewer is required. External calls send the supplied materials to that service and consume its account usage. Lite retains a final checklist and skips review.
A recording of the product questions, passage changes, reasoning, and review decisions.