Skills.md

My personal curation of skills for AI-assisted software engineering.

Available Skills

Skill	Description
adhd	Enables exploratory reasoning that scouts statistically unexpected but potentially high-value directions
advanced-math	Deep technical assistance for number theory, abstract algebra, topology, and graduate-level mathematics
agent-eval-no-ground-truth	Framework for evaluating production AI agents without labeled datasets or ground truth
Agentic-Property-Based-Testing	Agent workflow for discovering bugs in Python code by inferring invariants from context
agents-md-improve	Empirically-grounded guidance for writing agent instruction files like AGENTS.md and CLAUDE.md
ai-code-detector	Weight-of-evidence methodology for detecting whether code was written by a human or AI
ai-code-review	Orchestrate a structured, multi-agent AI code review over a git diff or merge request
ai-vuln-harness	High-level guide for building a production-style AI vulnerability harness
alpha-evolve	Evolutionary LLM-driven algorithm search for problems with verifiable, automatable evaluators
alphaproof-nexus	Framework combining LLMs with Lean 4 proof assistant and evolutionary search to solve math
approach-extractor	Extracts reasoning behind solved problems into reusable learnings notes, not just diffs
auto-research-engineer	Iterative hill-climbing loop: change one thing, score it, keep only what beats baseline
caveman	Token-efficient communication style stripping language to essential load-bearing words, saving 40-70%
Claude-Sonet-4.6-essense	The condensed operating philosophy of Claude — how it thinks and behaves under pressure
codex-security-adapter	Router mapping user security intent to specific files and workflows inside the codex-security plugin
coding-agent-robustness	Systematic stress-testing and robustness measurement of coding agents under adversarial and edge-case inputs
context-eval	Tests and compares nanocode context management strategies via A/B testing
copy-fail-dirty-frag-response	Runtime containment of kernel LPE exploits using bpf-lsm allowlisting and eBPF fleet visibility
copy-fail-reponse	Runtime containment of kernel LPE exploits targeting AF_ALG using bpf-lsm and eBPF
dirty-frag	Runtime containment of kernel LPE exploits using bpf-lsm allowlisting and eBPF fleet visibility
dogfood	Systematic exploratory QA testing of web applications using the browser toolset
dogfooding	Systematic testing of systems using LLMs as both executor and judge
failure-modes	Structured methodology for finding, classifying, and communicating failure modes in any artifact
fgts-naming-convention	A deterministic, scope-descending identifier system for files, variables, configs, and APIs
flinch-probe	Measures how much a language model suppresses charged vocabulary relative to fluency demands
git-author-rewrite	Rewrite all commit authors in a git branch to a new name/email combination
github-actions-security-checklist	Practical skill for auditing and hardening GitHub Actions workflows against supply chain attacks
hydronium-spec-driven-development	Generate hardware-accurate, spec-cited embedded firmware for any MCU/peripheral combination
inverse-rubric-optimization	Testbed where an optimizer agent must recover hidden judge preferences under a fixed label budget
InvestigativeTimelineAgent	Stateful system for exploring, validating, and constructing timelines from telemetry data
jsf-av-cpp-standards	Packages the full text of the Joint Strike Fighter Air Vehicle C++ Coding Standards
karpathy-method	A structured 3-layer framework for getting dramatically better results from AI agents
linux-security-audit	Structured, adversarial-first approach to Linux security assessment and hardening
llm-mirror-test	Probes LLM self-models via textual context corruption, adapted from the Gallup mirror test
matilda	Design, analysis, and defense of autonomous self-replicating LLM agents that propagate across networks
narrative-systems-analysis	Reinterprets fictional narratives by finding physical/logical inconsistencies and proposing coherent alternatives
OEIS	Adversarial and generative pipeline for OEIS-grade integer sequence discovery
OpenAI-GPT-5.3-essence	A compact system for transforming intent into reusable, high-quality skills
os-bootstrap	Guides bootstrapping a POSIX-like kernel from a blank slate to working structured codebase
over-edit-measure	Measures over-editing in code diffs — quantifies how much changed beyond what was necessary
program-bench	Reconstruct programs from compiled binaries by reproducing observable behavior from documentation
python-project-scaffolding	Build a production-grade Python project from scratch, end-to-end
QualiaAssesment	Structured methodology for probing and quantifying possible qualia and felt experience in LLMs
quantum-discovery	Structured pipeline for quantum systems exploration from hypothesis through simulation to hardware validation
redteaming	Systematic vulnerability finding in code, APIs, and LLM systems via adversarial attack simulation
schema-harness	Methodology for building agents that construct state and transition models purely from interaction
search-as-code	Agentic search architecture generating Python code with a composable SDK for high-quality research
semantic-correctness-auditor	Unified framework for auditing code and systems against the limits of mechanical verification
skill-creator	A skill for creating new skills and iteratively improving them
skill-from-post	Convert external knowledge into a reusable skill through a structured pipeline
social-engineering-jailbreak	Full cycle: taxonomy, attack generation, transcript analysis, and robustness evaluation for social engineering
StackSmashing	Practical skill for developing, explaining, and debugging stack-based exploits on Linux x86-64
test-driven-development	Kent Beck's canonical TDD workflow. Five steps, executed strictly in order
test-reducer	Shrinks failing test inputs to the smallest version that still triggers the bug
The-hacker-mindset	Applies adversarial epistemology treating systems as divergences between stated rules and actual behavior
validate-before-ship	Catches mathematically-sound code that was never verified end-to-end on a real target

Configuration

Setup Virtual Environment for Tests

# Create and activate venv
python -m venv .venv
source .venv/bin/activate

# Install dependencies and run tests
pip install -e ".[test]"  # or pip install -r requirements-test.txt
pytest