The Agent Skills Directory

[SAFE]: The skill is inherently defensive and aims to prevent supply-chain attacks. It provides specific instructions to verify the existence and integrity of dependencies against official registries without installing them.\n- [SAFE]: Contains strong procedural guardrails that enforce a 'plan-only' mode, explicitly forbidding commands like npm install or pip install during the audit phase to prevent accidental execution of malicious code.\n- [SAFE]: The skill addresses Indirect Prompt Injection risks by establishing clear boundary markers, such as the 'Plan only' guardrail and a phased approval process for any proposed changes to dependency manifests.

plan-dependency-provenance