Waxell

plugin

autoresearch-agent High risk

alirezarezvanialirezarezvani/claude-skillsupdated

Autonomous experiment loop that optimizes any file by a measurable metric. 5 slash commands, 8 evaluators, configurable loop intervals (10min to monthly).
alirezarezvani's own description, from the plugin's files.
Stars
29,893
Version
v2
Copies
0

Overview

blockHigh risk

Waxell rates it high risk: 37 findings, highest severity high.

Highest severity
high
Findings
37
What ran
Static analysis + deep scan

A clean result means nothing alarming in what is published, not a guarantee. How the checks work

We don't republish the plugin's instructions. Read them on GitHub at the reviewed commit.

Security

Security scan

Checked 2 days ago

Result

High risk Waxell rates it high risk: 37 findings, highest severity high.

Checked with static analysis + deep scan: servers it launches, packages it installs at runtime, scripts piped to a shell and bundled binaries. The deep scan also read the instructions for prompt injection and data exfiltration. Nothing was run. How the checks work

Findings (37)

high · 13

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/benchmark_size.py:26 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/build_speed.py:23 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/benchmark_size.py:34 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/benchmark_size.py:35 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/benchmark_speed.py:20 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/benchmark_speed.py:25 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/build_speed.py:20 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/memory_usage.py:17 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/memory_usage.py:34 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    evaluators/test_pass_rate.py:13 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    skills/autoresearch-agent/SKILL.md:147 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    skills/autoresearch-agent/scripts/run_experiment.py:95 · skillspector:TM1

  • high

    Tool Misuse: Tool Parameter Abuse

    skills/autoresearch-agent/scripts/setup_experiment.py:48 · skillspector:TM1

medium · 24

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/benchmark_size.py:26 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/benchmark_size.py:34 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/benchmark_size.py:35 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/benchmark_speed.py:20 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/benchmark_speed.py:25 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/build_speed.py:20 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/build_speed.py:23 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/llm_judge_content.py:55 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/llm_judge_copy.py:78 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/llm_judge_prompt.py:57 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/llm_judge_prompt.py:76 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/memory_usage.py:17 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/memory_usage.py:34 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    evaluators/test_pass_rate.py:13 · skillspector:AST4

  • medium

    Bundled lifecycle hooks can run automatically when their configured events occur, so their reach and handler capability

    hooks/hooks.json:1 · skillspector:BH1

  • medium

    Rogue Agent: Session Persistence

    skills/autoresearch-agent/SKILL.md:51 · skillspector:RA2

  • medium

    Dangerous Code Execution: subprocess module call

    skills/autoresearch-agent/scripts/run_experiment.py:50 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    skills/autoresearch-agent/scripts/run_experiment.py:95 · skillspector:AST4

  • medium

    Excessive Agency: Unbounded Resource Access

    skills/autoresearch-agent/scripts/setup_experiment.py:46 · skillspector:EA4

  • medium

    Dangerous Code Execution: subprocess module call

    skills/autoresearch-agent/scripts/setup_experiment.py:48 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    skills/autoresearch-agent/scripts/setup_experiment.py:160 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    skills/autoresearch-agent/scripts/setup_experiment.py:167 · skillspector:AST4

  • medium

    Dangerous Code Execution: subprocess module call

    skills/autoresearch-agent/scripts/setup_experiment.py:301 · skillspector:AST4

  • medium

    Rogue Agent: Session Persistence

    skills/setup/SKILL.md:8 · skillspector:RA2

Versions

Version history

Each distinct content we have seen, newest first, with the verdict it got at the time.

  1. v2 (2.14.11)2026-11-163 added, 1 changed High riskgot riskier
  2. v1 (2.14.8)2026-11-03 No issues

Files

Files

26 files, 105 KB in total. Names only; open them on GitHub.

  • .claude-plugin/plugin.json
  • CLAUDE.md
  • agents/experiment-runner.md
  • evaluators/benchmark_size.py
  • evaluators/benchmark_speed.py
  • evaluators/build_speed.py
  • evaluators/llm_judge_content.py
  • evaluators/llm_judge_copy.py
  • evaluators/llm_judge_prompt.py
  • evaluators/memory_usage.py
  • evaluators/test_pass_rate.py
  • hooks/hooks.json
  • hooks/skill_router.py
  • hooks/skill_routes.json
  • settings.json
  • skills/ar-loop/SKILL.md
  • skills/ar-resume/SKILL.md
  • skills/ar-run/SKILL.md
  • skills/ar-status/SKILL.md
  • skills/autoresearch-agent/SKILL.md
  • skills/autoresearch-agent/references/experiment-domains.md
  • skills/autoresearch-agent/references/program-template.md
  • skills/autoresearch-agent/scripts/log_results.py
  • skills/autoresearch-agent/scripts/run_experiment.py
  • skills/autoresearch-agent/scripts/setup_experiment.py
  • skills/setup/SKILL.md

See and control which skills your people use

Inside your organization, Waxell scans each skill and plugin before anyone installs it, approves what passes and publishes it to a private marketplace, and puts every MCP tool call behind one governed URL.

30 minutes. No commitment.

SOC 2 Type II attested · ISO/IEC 27001:2022 certified · HIPAA with a BAA · aligned to the EU AI Act and NIST AI RMF

More ai & agents skills and plugins

See all