arm-examples/llm-backend-scaffold
skills/llm-backend-scaffold/SKILL.md
Scaffold and integrate a new LLM backend/framework into this repository (new subdir under src/cpp/frameworks, wire LLM_FRAMEWORK routing, configuration options, downloads/models, tests, and docs). Use when adding a new framework, porting a backend, or refactoring backend glue code.
$ install --global
skillsauthnpx skillsauth add arm-examples/llm-runner llm-backend-scaffoldInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 23, 2026, 1:36 AM1.7s1 file scanned
SKILL.md
- name:
- llm-backend-scaffold
- description:
- Scaffold and integrate a new LLM backend/framework into this repository (new subdir under src/cpp/frameworks, wire LLM_FRAMEWORK routing, configuration options, downloads/models, tests, and docs). Use when adding a new framework, porting a backend, or refactoring backend glue code.
Add a backend (scaffold)
This repo selects a backend via LLM_FRAMEWORK at CMake configure time and routes to a backend-specific integration under src/cpp/frameworks/.
Windows note: if python3 isn’t available, use python (or py -3) for the scripts below.
Workflow
1) Choose the integration shape
- In-tree backend wrapper (preferred): implement a thin adapter that satisfies the project’s interface and build it from
src/cpp/frameworks/<new>/. - External dependency: if building from source, decide where it lives (usually as a subdirectory pulled into the build). Keep it deterministic and pinned.
If the backend needs downloads (models / jars / tools), plan to extend scripts/py/requirements.json and keep SHA256s accurate.
2) Create the backend directory
- Add
src/cpp/frameworks/<new-backend>/with aCMakeLists.txt. - Implement the backend adapter classes in that directory following the existing backend patterns.
Reference: skills/llm-backend-scaffold/references/backend-integration-checklist.md.
3) Wire CMake routing + configuration options
- Route the new backend in
src/cpp/frameworks/CMakeLists.txtbased onLLM_FRAMEWORK. - Update
scripts/cmake/configuration-options.cmaketo include the newLLM_FRAMEWORKvalue in the cache strings list. - If you need new knobs, add them as cache variables/options in a backend-scoped CMake file and document them in
README.md.
4) Add a model configuration file
- Add at least one JSON config under
model_configuration_files/for the new backend. - Ensure the config format is supported by the shared config code in
src/cpp/config/.
5) Ensure tests exercise it
- Update
test/CMakeLists.txtto include at least one config file in theCONFIG_FILE_NAMEselection for the new framework (mirrors existingelseif (${LLM_FRAMEWORK} STREQUAL "...")cases). - Build and run:
cmake --preset=native -B build -DLLM_FRAMEWORK=<new>→cmake --build ./build --parallel→ctest --test-dir ./build --output-on-failure
6) Update downloads (if needed)
- Add model/tool entries to
scripts/py/requirements.jsonunder the appropriate section. - Compute and set
sha256sumfor every new URL. - Remember: configure runs downloads via
scripts/cmake/download-resources.cmake→scripts/py/download_resources.py; gated models requireHF_TOKEN(or~/.netrc).
7) Document the backend
- Update
README.md(supported frameworks, build flags, supported models, runtime notes). - Add troubleshooting notes to
TROUBLESHOOTING.mdif the backend has platform-specific gotchas.
Related Skills
arm-examples/llm-update-downloads
tools
VerifiedTrustedCommunity
Update scripts/py/requirements.json entries (URLs + sha256sum) for models/tools, validate hash changes, and keep downloads deterministic without committing artifacts. Use when adding or refreshing model/tool downloads.
4SKILL.mdUpdated Apr 4, 2026
arm-examples/llm-session-start
tools
VerifiedTrustedCommunity
Run fast “session start / doctor” checks for this repository (toolchain + wiring sanity, framework version report, optional upstream update check), optionally generate a debug bundle, and when needed bump pinned backend framework versions with build+ctest verification. Use at session start or when upgrading llama.cpp/onnxruntime-genai/mediapipe/mnn pins.
4SKILL.mdUpdated Apr 4, 2026
arm-examples/llm-jni-smoke
tools
VerifiedTrustedCommunity
Run a fast JNI-focused build/test smoke check (JNI on, minimal test run), and isolate JNI toolchain issues. Use when changing JNI/Java code or validating JNI setup.
4SKILL.mdUpdated Apr 4, 2026
arm-examples/llm-debug-test-failures
development
VerifiedTrustedCommunity
Debug failing LLM integration tests caused by model output drift, incorrect context/runtime parameters (contextSize, batchSize, threads), prompt/template mismatches, or backend/framework regressions. Use when tests fail and you need to see the model response, reproduce a single failing CTest, or trace issues into src/cpp/frameworks (llama.cpp, onnxruntime-genai, mediapipe, mnn).
4SKILL.mdUpdated Apr 4, 2026