Atlas / MCP servers / rrrrrredy / research-toolkit

research-toolkitCAUTION

Name: research-toolkit
Author: rrrrrredy

mcp/rrrrrredy/research-toolkit

MCP server by rrrrrredy · agent plugins · agent skill · ai research · Audited 2026-10-13

Research methods, workflow tools, and independent review for AI agents, packaged as a Skill, plugin, and local MCP server.

Verdict

CAUTION

Grade

Trust score

77 /100

Exposed tools

7 4r · 3w · 0d

Transport

stdio

License

MIT

Stars

204

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

[](./LICENSE) [](https://github.com/rrrrrredy/research-toolkit/actions/workflows/framework-checks.yml)

English · 简体中文

Existing user and development tasks are independent tasks, and tests actually executed are real tests. Development use, publication and reuse do not change task identity; distinguish tasks by their research questions. Preserve the supplied sources and data, and report execution counts and outcomes from actual run records.

Help AI agents write research reports with concrete judgments, traceable sources, and explicit review choices.

Research methods · Callable workflow tools · Review choices. For industry research, product comparisons, company analysis, and technical research.

Quick start · Use with your agent · Research case · Evaluation results · Lite quickstart · Documentation

Quick start

Start with `lite`. Allow about 5–15 minutes with your agent ready to use. Copy this request; the topic and two fictional sources are supplied:

Use https://github.com/rrrrrredy/research-toolkit and read SKILL.md.
Choose profile=lite. Read examples/lite/start.en.json for the brief
and examples/lite/sources.md for the source notes.
Recommend a helpdesk plan for the six-person team in 350–500 English
words. Use only the supplied evidence, register sources and claims,
and draft section by section. Skip research_review and Full delivery
gates; complete all six Lite checklist items with concrete evidence.
Return the short report, separate source/claim records, and checklist.
Disclose the fictional evidence and absence of independent review.

You receive a **short report + sources/claims trail

Read from source at commit 27e4d01246ecOBSERVED · 2026-10-13

Exposed tools (7)

4 read · 3 write · 0 destructive.

Tool	Risk	Description
`research_check_draft`	read	Check draft records and declared whole-artifact length before review or finish, without model calls.
`research_check_reviewer`	read	Check trusted reviewer configuration locally, without sending report data or paid model calls.
`research_finish`	write	Verify report delivery and publish its receipt/terminal state only when checks pass.
`research_guide`	read	Read brief/collect/analyze/draft/review/revise/final methods without a task.
`research_review`	write	Run the configured content reviews and any declared audits, retaining original evidence.
`research_start`	write	Create a research task or complete its missing brief. Returns its directory and stage guidance.
`research_status`	read	Read progress and stage methods. Optional stage advances to collect/analyze/draft/revise only

Trust audit

CAUTIONgrade C · trust 77/100 Install with care. The audit found things worth knowing before you trust its output.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	WARN
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	PASS
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (2 observation(s))
Network: declared (10 observation(s))
Shell: none-observed
Dependencies: pinned
Secrets in source: none-found

Findings (25)

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

evals/studies/2026-10-10-lite-four-models/run_live_study.py:99

config={'schema_version':1,'profile':'lite','source_access':'fixed_supplied_sources','cohort_status':'prospective_development','publication_rights_confirmed':True,'selection_statement':'Five purposive

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

scripts/check_blinded_evals.py:150

settings = {**self.config["author"], "base_url": f"http://127.0.0.1:{server.server_port}/v1"}

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

scripts/check_profiles_reviewers.py:470

config = {"reviewers": [{"id": "http", "backend": "generic", "model": "synthetic", "base_url": f"http://127.0.0.1:{server.server_port}/v1"}]}

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

scripts/check_profiles_reviewers.py:485

"http://127.0.0.1:8000/v1", "http://[::1]:8000/v1"):

MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319

scripts/check_profiles_reviewers.py:488

for url in ("http://review.example/v1", "http://10.0.0.1/v1",

MEDIUMObfuscation / stealth · obf.base64_blob · CWE-506, CWE-94

evals/diagnostics/2026-10-12-astra-calibration/code/check_judge_calibration.py.b64:1

IyEvdXNyL2Jpbi9lbnYgcHl0aG9uMw0KIiIiUmVhZC1vbmx5IGNoZWNrcyBmb3IgcHVibGljIGp1ZGdlIGNvbnRyb2xzOyBubyBtb2RlbCBjYWxscyBvciBxdWFsaXR5IHZlcmRpY3RzLiIiIg0KaW1wb3J0IGFyZ3BhcnNlDQppbXBvcnQgY29weQ0KaW1wb3J0IGhh

MEDIUMObfuscation / stealth · obf.base64_blob · CWE-506, CWE-94

evals/diagnostics/2026-10-12-astra-calibration/code/codex_judge.py.b64:1

IyEvdXNyL2Jpbi9lbnYgcHl0aG9uMwoiIiJGcmVzaCwgdG9vbC1kaXNhYmxlZCBDb2RleCBqdWRnZSB0cmFuc3BvcnQuIENMSSBtb2RlbCBpZGVudGl0eSBpcyByZXF1ZXN0ZWQgb25seS4iIiIKaW1wb3J0IGpzb24KaW1wb3J0IG9zCmZyb20gcGF0aGxpYiBpbXBv

MEDIUMObfuscation / stealth · obf.base64_blob · CWE-506, CWE-94

evals/diagnostics/2026-10-12-astra-calibration/code/run_blinded_evals.py.b64:1

IyEvdXNyL2Jpbi9lbnYgcHl0aG9uMw0KIiIiUHJlcGFyZSBvciBydW4gYSBzZWVkZWQsIGZpeGVkLXNvdXJjZSwgaW5zdHJ1Y3Rpb24tb25seSBMaXRlIHBhaXJ3aXNlIHBpbG90LiIiIg0KZnJvbSBfX2Z1dHVyZV9fIGltcG9ydCBhbm5vdGF0aW9ucw0KaW1wb3J0

MEDIUMObfuscation / stealth · obf.base64_blob · CWE-506, CWE-94

evals/diagnostics/2026-10-12-astra-calibration/code/run_judge_calibration.py.b64:1

IyEvdXNyL2Jpbi9lbnYgcHl0aG9uMwoiIiJGcmVlemUgYW5kIGV4ZWN1dGUgcHVibGljIGNhbGlicmF0aW9uIGNvbnRyb2xzIGluIGZyZXNoIENvZGV4IGp1ZGdlIGNvbnRleHRzLiIiIgppbXBvcnQgYXJncGFyc2UKZnJvbSBkYXRldGltZSBpbXBvcnQgZGF0ZXRp

MEDIUMObfuscation / stealth · obf.base64_blob · CWE-506, CWE-94

evals/diagnostics/2026-10-12-longcat-format/driver.py.b64:1

IiIiTmV3IGZyb3plbiBMb25nQ2F0IGZvcm1hdCBwcm9iZXM7IG5ldmVyIHNhbHZhZ2Ugb3IgcmVzY29yZSBkZXZlbG9wbWVudCBmYWlsdXJlcy4iIiIKZnJvbSBkYXRldGltZSBpbXBvcnQgZGF0ZXRpbWUsIHRpbWV6b25lCmltcG9ydCBnZXRwYXNzCmltcG9ydCBo

LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94

scripts/check_development_evidence.py:28

h.require(h.sha(base64.b64decode((root / (name + ".py.b64")).read_bytes().strip(), validate=True)) == plan[name + "_sha256"], "Format code archive mismatch")

LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94

scripts/check_feedback_probes.py:33

raw = base64.b64decode(safe(root, name + ".b64").read_bytes().strip(), validate=True)

LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94

scripts/check_live_diagnostics.py:54

raw = base64.b64decode(local(root, "code/" + name + ".b64").read_bytes().strip(), validate=True)

LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94

scripts/check_research_workflow.py:912

self.assertEqual(base64.b64decode(result["capture"]["raw_output_base64"]["stderr"]), b"warning: \xff")

LOWObfuscation / stealth · obf.decode_call · CWE-506, CWE-94

scripts/check_research_workflow.py:922

self.assertEqual(base64.b64decode(capture["raw_output_base64"]["stdout"]), b"broken \xff")

LOWPrompt injection · prompt.transfer_instruction · CWE-94, CWE-1427

docs/case-study/api-or-browser/reviews/01-draft-review.html:71

<p id="L58"><strong>S08: <a href="https://docs.github.com/en/apps/creating-github-apps/registering-a-github-app/choosing-permissions-for-a-github-app">GitHub: Choosing permissions for a GitHub App</a>