Report · marked 25 Aug 2026
ggml-org/llama.cpp
Solid foundations, held back in a few specific places.
C++ · 125,628 stars · 3,498 files · branch master
74
/100
B+
Instructions
23/27Whether the repo tells an agent how to behave before it starts guessing. Losing marks on instruction quality.Setup
9/18Whether an agent can install the project and get it running without a human. Losing marks on deterministic install and environment config.Verification loop
20/22Whether an agent can check its own work. This is the category that most decides whether agent output is trustworthy. Losing marks on tests exist.Context economy
11/20Whether the repo fits in a context window, or fights it. Losing marks on file sizes fit in context and repository weight.Navigability
9/10How quickly anything new can orient itself. Losing marks on repository description.
Higher is better. Accent marks what is costing you most.
Fix these first
01
File sizes fit in context
A single file that fills the context window forces an agent to work from fragments, and it will confidently edit code it never saw. Splitting the worst offenders pays for itself immediately.
02
Environment config
Add a .env.example listing every variable with a safe placeholder value. It's the cheapest possible fix and it unblocks the whole first run.
03
Deterministic install
Commit the lockfile your package manager generates. Without it, an agent's install and your install are different builds, and 'works on my machine' becomes unfalsifiable.
The full marking
Every deduction below names the file or setting it came from.
Instructions 23/27 A
Whether the repo tells an agent how to behave before it starts guessing.
✓Agent instruction file
12/12
AGENTS.md is present, which is the format the most tools read.
Also found: CLAUDE.md
✗Instruction quality
8/12
Real instructions, with gaps.
Worth fixing: it never states a build, test, or run command.
12,162 characters, a workable length
uses headings, so an agent can skim it
includes copyable code blocks
✓README as an entry point
3/3
README has a clear getting-started section.
Setup 9/18 C+
Whether an agent can install the project and get it running without a human.
✗Deterministic install
3/6
Only requirements.txt, which pins loosely at best.
Commit the lockfile your package manager generates. Without it, an agent's install and your install are different builds, and 'works on my machine' becomes unfalsifiable.
✓Discoverable commands
6/6
Commands are declared where an agent will look for them.
Makefile
pyproject.toml scripts
scripts/ directory
✗Environment config
0/4
Nothing documents the environment this needs. An agent will get a runtime error it can't diagnose.
Add a .env.example listing every variable with a safe placeholder value. It's the cheapest possible fix and it unblocks the whole first run.
✗Pinned runtime version
0/2
No pinned runtime version.
Add a .nvmrc, .python-version, or equivalent so the agent's toolchain matches yours.
–Reproducible environment
n/a
Not applicable — a library doesn't need a container to be worked on.
Verification loop 20/22 A
Whether an agent can check its own work. This is the category that most decides whether agent output is trustworthy.
✗Tests exist
6/8
243 test files against 2144 source files.
e.g. examples/llama.android/lib/src/test/java/android/llama/cpp/ExampleUnitTest.kt, examples/sycl/win-test.bat, ggml/src/ggml-opencl/kernels/gemv_noshuffle_q4_0_f32_spec.cl
✓Test command is discoverable
7/7
An agent can find and run `pytest`.
✓Continuous integration
4/4
50 GitHub Actions workflows define what "passing" means.
✓Lint and format rules
3/3
.flake8 encodes the house style.
–Static type checking
n/a
Not applicable — the C++ compiler type checks every build.
Context economy 11/20 B-
Whether the repo fits in a context window, or fights it.
✓No committed build output
6/6
No generated directories committed.
✓.gitignore hygiene
3/3
.gitignore covers 102 patterns.
✗File sizes fit in context
0/6
60 source files are over 100KB.
A single file that fills the context window forces an agent to work from fragments, and it will confidently edit code it never saw. Splitting the worst offenders pays for itself immediately.
vendor/miniaudio/miniaudio.h — 4012KB
ggml/src/ggml-opencl/ggml-opencl.cpp — 1181KB
ggml/src/ggml-vulkan/ggml-vulkan.cpp — 997KB
✗Repository weight
2/5
About 434MB checked out. Large enough that cloning and searching are both slow.
Large binaries and vendored trees slow every operation an agent performs. Git LFS or a separate assets repo keeps the working tree navigable.
Here is your AGENTS.md
Drafted from what is actually in this repository: the install
command from your lockfile, the commands you already declare, your real directory layout. Anything
marked TODO needs a person. Save it at the root as AGENTS.md.
# AGENTS.md LLM inference in C/C++ ## Setup ``` pip install -r requirements.txt ``` ## Commands ``` make https ``` ## Layout - `tools/` 922 source files - `ggml/` 576 source files - `src/` 214 source files - `examples/` 99 source files - `conversion/` 89 source files - `common/` 70 source files ## Conventions - Tests live alongside the code they cover, following `examples/llama.android/lib/src/test/java/android/llama/cpp/ExampleUnitTest.kt`. - CI defines what passing means. See `.github/workflows/ai-issues.yml`, and keep it green. - TODO: add the two or three conventions a newcomer always gets wrong here. ## Gotchas - `vendor/miniaudio/miniaudio.h` is 4012KB. It will not fit comfortably in context, so read it in parts. - No lockfile is committed, so an install here may not match what CI produced. --- Drafted by agentreadme.com from what is in this repository. Everything marked TODO needs a human. Check it in as AGENTS.md at the root.
Open the raw markdown
or
curl -o AGENTS.md agentreadme.com/draft/ggml-org/llama.cpp.md
Show the mark
The badge re-checks daily, so it keeps up as the repository changes. Use mark again to force it now.
[](https://agentreadme.com/ggml-org/llama.cpp)Other C++ repositories, marked
Think this mark is wrong?
Every deduction above names the file it came from, so this can be settled by looking. If a check missed something, that is a rule worth fixing.