Agent Skills spec · MIT · CI-green

Low-level engineering knowledge that verifies itself

Works with Claude Code · Cursor · GitHub Copilot · VS Code · Gemini CLI · OpenCode

0
skills
0
domains
0
source-backed
0
primary sources
0
traced claims
explore

Search the entire skill library

Every skill is a compact SKILL.md: when to use, what the agent often gets wrong, how to reason correctly, and how to verify.

why

What the agent often gets wrong

Named failure classes, catalogued from real incidents — not vibes.

Assembly hallucinations

Invented mnemonics that do not exist (movqad), AT&T/Intel operand inversion, zeroed immediates.

Fake parallelism

Thread-safe-looking code running on one thread. The gate measures real thread counts and wall-clock scaling.

Crate hallucination

Nonexistent crates that resemble real ones — a typosquatting risk (5.2–21.7% of generated packages).

Misleading verification

A "passing" harness that does not test the target. Fixed-shape oracles certify buggy GPU kernels.

research

Grounded in 55+ documented failures

Three source-traced surveys turned real AI-agent failures into an engineering playbook.

install

Load the whole pack in seconds

git clone https://github.com/TrothByte/low-level-skills-trothbyte

quality

Verification is executed, not asserted

Source-backed

65 of 124 skills were verified by executing examples on real toolchains: GCC 16.1, rustc 1.97, GDB, objdump, CMake/Ninja.

Claim → source → skill

Every normative claim is traced in registry/claims.yaml, backed by 177 primary sources.

Honest stability

Where a toolchain is unavailable, skills are marked researched with the exact verification command.

One gate for everything

python tools/validate.py checks all 124 skills, the registry, and every source trace in seconds.