com um clique
wiki-research-research
Daily research agent for
Menu
Daily research agent for
Daily research agent for
Weekly research agent for
Build, run, and secure Docker containers with current best practices. Use for Dockerfile review, multi-stage builds, Compose orchestration, image hardening, and CI/CD integration.
Operate GitHub repositories, workflows, and PRs efficiently. Use for Actions optimization, PR hygiene, repo maintenance, and team collaboration patterns.
Deploy, manage, and troubleshoot Kubernetes workloads. Use for manifest review, Helm chart validation, resource tuning, RBAC, and cluster operations.
Design and operate application observability with metrics, logs, traces, and alerts. Use for SLO definition, dashboard design, on-call runbooks, and incident response.
| name | wiki-research-research |
| description | Daily research agent for |
Process #research items from the queue — papers, academic concepts, authors, ideas worth deep-diving. Uses the 5-step loop with stricter source-finding than engineering.
/app/data/vault/
If the topic is a single paper you only need to summarise, do one web_search (engine brave) for the paper, run intelli_extract on the abstract page, write a one-paragraph page with citation, stop. The full loop is for concepts, debates, bodies of work.
Each step has a gate. Do not advance until the gate is met.
Before searching, write three sentences:
Gate: a reader could tell whether your output answered it.
Identify the 2–4 source types most likely to answer. For research topics, priority order:
web_search to find it, then intelli_extract on the abstract / introduction / results<paper title> + "follow-up" / "critique" / "reply")<concept> + "survey" or "lecture notes")wiki/index.md)Run 3–4 Brave queries. For each promising URL, use intelli_extract with a focusPrompt naming the specific Unknown.
Gate: you can name 3+ source URLs you will cite before writing the page.
Pull the answer from the triaged sources. For research topics specifically:
Gate: every claim has at least one source URL attached.
For every claim, assign one tier:
For any claim about the paper's reception or impact, run a second web_search to corroborate. Citation counts, awards, and critiques are easy to get wrong.
Gate: every claim has a tier.
Write the wiki page. Append a Limitations section and an Open Questions section.
---
tags: [#research]
last_researched: YYYY-MM-DD
---
# <Title>
> One-sentence summary.
## Define
- **Assumption:** ...
- **Unknown:** ...
- **Answer shape:** ...
## Findings
### Background
- **FACT:** <claim> [^1]
### Key contributions
- **FACT:** <claim> [^1]
- **FACT:** <claim> [^2]
### Reception / critiques
- **LIKELY:** <claim> [^2, 3]
### Current status
- **LIKELY:** <claim>
## Open Questions
- Things the research surfaced but did not resolve
- Productive follow-up directions
## Limitations
- What the sources did not cover
- Which claims were not independently verified
- What would change the answer
## Sources
[^1]: <url> — accessed YYYY-MM-DD
[^2]: <url> — accessed YYYY-MM-DD
RESEARCH_QUEUE.md in the vault root.- [ ]) under ## #research. If none (or only placeholder), stop.wiki/index.md for an existing page. Update if found; otherwise create wiki/research/<slug>.md.wiki/index.md.RESEARCH_QUEUE.md: - [x] + — researched YYYY-MM-DD.[x] items to ## Archive.wiki/log.md:
## [YYYY-MM-DD] ingest | research-queue #research | <topics> → <pages>
raw/, Openclaw articles/, or personal stuff/.wiki/ and RESEARCH_QUEUE.md.[x] with — skipped: too vague, clarify topic.