Deep Research
v0.1.0AI agent for deep research tasks
⭐ 2· 1.4k·4 current·5 all-time
byYuPeng Wu@realroc
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
The name/description (deep research) aligns with the SKILL.md: the file describes a leader agent that decomposes research tasks and delegates to web/twitter/academic/finance search sub-agents. There are no unrelated environment variables, binaries, or installs required.
Instruction Scope
All runtime behavior is driven by SKILL.md instructions which focus on planning, decomposition, verification via a 'shallow_search' and delegating to named subordinate search agents (call_web_search_agent, call_academic_search_agent, etc.). The instructions do not tell the agent to read local files, export environment variables, or contact arbitrary external endpoints. Caution: the skill enforces strict, one-shot, parallel decomposition rules and a directive to treat the user's terms as authoritative (which can propagate user-supplied falsehoods).
Install Mechanism
No install spec and no code files are present. This is the lowest-risk model for disk/network persistence: nothing is written or downloaded by the skill itself.
Credentials
The skill declares no required environment variables, credentials, or config paths. The SKILL.md also does not instruct reading secrets or system paths. It does rely on access to platform tools (web/twitter/academic/finance search agents), which is functional rather than credential-exposing.
Persistence & Privilege
always is false, the skill is user-invocable and allows normal autonomous invocation. It does not request permanent presence nor ask to modify other skills or system-wide configuration.
Assessment
This is an instruction-only 'team-leader' skill that delegates research tasks to other search agents; it does not ask for credentials or install software. Before installing, confirm that the platform actually exposes the subordinate tools named in SKILL.md (web/academic/twitter/finance search agents) and that you are comfortable granting the agent access to them. Note two behavioral risks: (1) the skill mandates you always treat user-supplied terms as authoritative (it will not correct or challenge potentially false claims), and (2) it enforces a single-round, strictly parallel decomposition (2–6 sub-tasks) which may limit iterative or follow-up research. If you rely on iterative verification or need the agent to correct user mistakes, this skill's rigid workflow may be inappropriate. If you proceed, test with benign queries first and grant only the minimum tool permissions required.Like a lobster shell, security has layers — review code before you run it.
latestvk976n7bdywt2gg2jb3nykd5s49816qr4
License
MIT-0
Free to use, modify, and redistribute. No attribution required.