The Agent Skills Directory

[SAFE]: The skill is a meta-utility for developing other skills. It provides structured guidance on creating checkpoints.yaml files, including mechanical checks (regex, file existence) and LLM-based reviews.
[SAFE]: It explicitly warns against and provides logic to prevent 'Vendor Leakage' (accidentally scanning node_modules, vendor/, or .Build/ directories), which is a best practice for security and performance.
[SAFE]: The workflow involves running validation scripts located in ${CLAUDE_PLUGIN_ROOT}. These are platform-provided paths for internal tooling and do not represent external remote code execution in a malicious context.
[SAFE]: The skill includes instructions to avoid 'Skill-relative script paths,' which prevents errors where a skill tries to execute local scripts in a target repository where they don't exist, effectively mitigating a class of broken execution path issues.
[SAFE]: All referenced URLs and repositories (github.com/netresearch/*) belong to the verified author 'netresearch', representing legitimate vendor resources.

add-checkpoints