Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Paper Research Agent
v1.0.0Autonomous multi-agent paper research system. When user wants to research a topic, find related papers, or analyze academic literature, use this skill to orc...
⭐ 0· 115·0 current·0 all-time
by崔之行@changer-changer
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The name/description align with the code and instructions: it searches arXiv, downloads PDFs, and coordinates per-paper analyses. However the skill assumes the presence of an external 'paper-reader' skill/tool at a hard-coded path (~/.openclaw/skills/paper-reader/read_paper.py) and suggests a launch_agents.py script that is not present in the bundle. Those undeclared dependencies are unexpected and weaken the declared self-contained purpose.
Instruction Scope
Runtime instructions perform network downloads and write PDFs and task files to the workspace (expected), but they also instruct the agent to 'spawn as many agents in parallel as possible' using sessions_spawn. That gives the skill broad discretion to create many sub-agents (resource exhaustion or wide action surface). The SKILL.md also runs subprocess commands and expects external tools; it references other skill paths and an absent launch script, giving ambiguous/incomplete guidance.
Install Mechanism
There is no formal install spec in the registry, but the bundled script auto-installs Python packages via pip at runtime (arxiv, requests, pdfplumber). Auto-pip-install is common but increases risk because it executes package installation from PyPI during execution rather than a reviewed install step. This is moderate risk (supply-chain / arbitrary code from PyPI) and should be considered when running in production.
Credentials
The skill requests no secrets or environment variables (good). It does, however, access and write files under the agent workspace and references other skills' paths (~/.openclaw/skills/paper-reader), which is not declared. That implies implicit reliance on other skill artifacts and file-system access that the description doesn't call out explicitly.
Persistence & Privilege
The skill is not forced-always and allows normal autonomous invocation. The main privilege concern is operational: instructions to spawn many parallel sub-agents can amplify the blast radius of any misbehaving sub-agent. There is no sign the skill modifies other skills' configs, but it does read/write workspace files and produce tasks for autonomous agents.
What to consider before installing
Before installing or running this skill:
- Inspect scripts/research_pipeline.py and references/analysis_standards.md yourself to verify behavior and confirm there are no hidden network endpoints or unexpected commands.
- Note it will pip-install packages at runtime (arxiv, requests, pdfplumber). If you need supply-chain assurance, pre-install vetted versions or run in an isolated environment.
- The skill expects a paper-reader tool at ~/.openclaw/skills/paper-reader/read_paper.py and references a launch_agents.py that is not present; ensure those dependencies exist and are trustworthy.
- Limit parallelism: do not launch 'as many agents as possible' on your machine—test with a small max_papers and controlled concurrency to avoid resource exhaustion or runaway agent spawning.
- Run first in a sandbox or restricted environment (network and process limits) and review generated _agent_tasks.json and task files before actually invoking sub-agents.
- If you lack the ability to audit the skill code, treat it as higher-risk and prefer manual execution of its components rather than fully autonomous runs.Like a lobster shell, security has layers — review code before you run it.
latestvk973j1sx8y287wbvcxq1jtrwb9834w5h
License
MIT-0
Free to use, modify, and redistribute. No attribution required.