OpusFlame Deep Research
PassAudited by ClawScan on May 1, 2026.
Overview
This instruction-only skill is coherent for deep research, but it will automatically use multiple model agents, web tools, and persistent research notes when invoked.
Install only if you want an autonomous, multi-model research workflow. Do not use it with confidential topics unless you are comfortable with the prompt being sent to several model services and with intermediate research notes being saved.
Findings (3)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
A single deep-research request may generate many model calls and web requests, which can consume resources and share the research topic with tool/model services.
Invoking the skill starts a chained autonomous workflow using sub-agent, search, and fetch tools. This is disclosed and central to the deep-research purpose, but users should notice the breadth of automated tool use.
Spawn 4 sub-agents using `sessions_spawn`, each with a different model ... Use web_search extensively (minimum 10 unique searches) ... Use web_fetch to read full articles
Use it for genuinely deep research tasks, and ask the agent to reduce model count or search breadth for low-stakes or sensitive topics.
If the research question contains confidential business, personal, or investment information, it may be exposed to multiple model contexts.
The skill intentionally distributes the same research task across multiple model agents/providers. That is aligned with its multi-model purpose, but the artifacts do not describe privacy boundaries between those agents.
Model 1: gemini (google/gemini-2.5-pro) ... Model 2: o3 (openai/o3) ... Model 3: opus (anthropic/claude-opus-4-6) ... Model 4: minimax (minimax/MiniMax-M2.5)
Avoid submitting sensitive or proprietary details unless you are comfortable with the involved model services; redact private information where possible.
Research topics, sources, and intermediate conclusions may remain available after the task and could be reused or over-trusted later.
The skill stores research outputs in a memory path. This is disclosed and useful for research continuity, but it creates persistent records derived from model outputs and web sources.
Save each model's output: memory/research/[topic]-gemini-[date].md ... memory/research/[topic]-o3-[date].md ...
Review saved research notes for accuracy and sensitivity, and delete or avoid saving outputs for confidential topics.