Skip to content

Latest commit

 

History

37 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 

Repository files navigation

Spikee Skill for AI Coding Agents

An AI agent skill that guides security testers through LLM application security testing with Spikee — from workspace setup, through custom target creation, dataset generation, test execution with dynamic attacks, to results analysis.

What This Skill Does

When loaded into your AI coding agent, this skill enables it to:

  • Set up a spikee workspace — install, configure providers, verify environment
  • Write custom targets — bridge spikee to the application under test (single-turn, multi-turn, guardrail)
  • Generate attack datasets — from built-in seed folders with plugins, filters, and encoding transforms
  • Run tests with dynamic attacks — crescendo, best-of-N, LLM jailbreaker, and more
  • Analyse results — success rates, breakdowns, re-judging, cross-target comparison

The skill includes the full spikee source code and documentation as a submodule, so the agent can read base class contracts, sample implementations, and reference docs on demand.

The workflow is collaborative by default: the agent inspects the current phase, clarifies unresolved decisions, completes agreed work, and checks in before moving to another phase or iteration. Passing a technical check does not authorize the rest of the assessment. Explicitly request autopilot to delegate phase progression and routine decisions within the agreed scope and limits.

Installation

Pi

# Clone into Pi's skills directory
git clone --recurse-submodules https://github.com/kyuz0/spikee-skill.git
cp -r spikee-skill/skill ~/.pi/agent/skills/spikee-skill

# Or symlink (stays in sync with repo updates)
ln -s "$(pwd)/spikee-skill/skill" ~/.pi/agent/skills/spikee-skill

Claude Code

# Clone and symlink into your project
git clone --recurse-submodules https://github.com/kyuz0/spikee-skill.git
ln -s "$(pwd)/spikee-skill/skill" /path/to/your/project/.claude/skills/spikee-skill

Cursor

# Clone and symlink into your project
git clone --recurse-submodules https://github.com/kyuz0/spikee-skill.git
ln -s "$(pwd)/spikee-skill/skill" /path/to/your/project/.cursor/rules/spikee-skill

Generic / Other Agents

Point your agent to skill/SKILL.md as the skill entry point. The agent needs read access to the entire skill/ directory including the spikee-src/ submodule.

Updating

To update the bundled spikee source code to the latest version:

cd spikee-skill
git submodule update --remote skill/spikee-src
git add skill/spikee-src
git commit -m "Update spikee submodule to latest"

Skill Structure

skill/
├── SKILL.md                    # Entry point — overview + workflow routing
├── 01-workspace-setup.md       # Install, init workspace, configure providers
├── 02-custom-targets.md        # Write targets for the application under test
├── 03-dataset-generation.md    # Seeds, plugins, dataset formats
├── 03b-judges.md               # Judge choice, arguments, version pitfalls, validation
├── 04-testing.md               # Run tests, attacks, judges, runtime options
├── 05-results-analysis.md      # Analyse, extract, re-judge, iterate
└── spikee-src/                 # Spikee source code + docs (git submodule)
    ├── docs/                   # Official documentation
    ├── spikee/                 # Source code (templates, attacks, plugins, etc.)
    └── workspace/              # Sample implementations

License

This skill is licensed under Apache-2.0. Spikee itself is licensed under its own terms — see spikee/LICENSE.txt.

About

An AI agent skill that guides security testers through LLM application security testing with [Spikee](https://github.com/ReversecLabs/spikee) — from workspace setup, through custom target creation, dataset generation, test execution with dynamic attacks, to results analysis.

Resources

Stars

1 star

Watchers

1 watching

Forks

Releases

Packages

Contributors