Ctx research
Agent skills I use day to day (lowband / handoff / multi-agent-driven / ctx-* family)
npx -y skills add keli-wen/skills --skill ctx-researchAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
3 things to look at
- 20 days oldThe repository was created 20 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Deep research grounded in the project's durable context system. Use for "ctx research", "基于 context 调研 X", "结合我们的 contexts 做深度研究". Reads existing contexts first, sharpens the question to the real knowledge gap, fans out multi-agent research, verifies key claims, and files results back.
SKILL.md
2.2 KB, 396 tokens by cl100k_base, as published. Nobody here has run it
ctx-research
Research that starts from what the project already knows and ends with the project knowing more. It differs from generic deep research in two ways: the question is sharpened against existing contexts before any searching, and verified findings are written back into the context system.
Language: mirror the user's language.
Steps
- Ground: locate the project's durable context layer (e.g. a
contexts/tree with a map file, or a docs/ADR system). Read the entries relevant to the question. Sort what you find into: already answered (do not re-research), assumed but unverified, genuinely open. - Refine: restate the research question as the delta between what the user needs and what contexts already hold. Confirm scope with the user only if the delta is ambiguous; otherwise proceed.
- Fan out (using
multi-agent-drivenpatterns): parallel researchers with distinct angles — official docs, community and prior art, papers, code in the wild. Each returns raw, citable findings with source name and URL or path. - Verify: every claim that will drive a decision gets an adversarial pass — a separate check that tries to refute it or show it is stale. Mark key claims confirmed / plausible / refuted.
- Synthesize: one report — answer first, then evidence, then open questions. Keep cited facts and your own judgment visibly separate.
- File back: hand the durable part to
to-ctx, which places it in the project's context system per its rules. The chat gets the summary plus the file path.
Rules
- Never re-research what contexts already answer; cite the context file instead.
- Verification effort scales with decision impact, not with how interesting a claim is.
- Dependency direction: may use
multi-agent-driven,to-ctx, andhandoff; never invokes user-facing skills.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.