The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and follow behavior-modifying instructions from SKILL.md files located within the node_modules directory of the project.
Ingestion points: Instructions are read from @fastgpt-plugin/cli/skills/cli-usage/SKILL.md and multiple paths under @fastgpt-plugin/sdk-factory/skills/ using sed commands.
Boundary markers: None present; the skill explicitly directs the agent to treat these external files as the "source of truth" for plugin implementation.
Capability inventory: The skill provides access to powerful tools including git, gh (GitHub CLI), pnpm package manager, and node execution.
Sanitization: No validation or sanitization is performed on the content of these external files before they are read into the agent's context.
[COMMAND_EXECUTION]: The skill provides instructions for standard development operations within the labring repository ecosystem.
Evidence: Commands include git clone, gh repo fork, pnpm install, and npx @fastgpt-plugin/cli check.
These actions are consistent with the skill's primary purpose of setting up a coding environment and managing plugin dependencies.
[DYNAMIC_EXECUTION]: The skill uses a dynamic Node.js script to locate package roots at runtime.
Evidence: A node -e script uses import.meta.resolve and file system checks to find the installation path of the @fastgpt-plugin/sdk-factory package.
This logic is used to resolve relative paths for loading documentation and does not incorporate untrusted external input into the execution string.

develop-fastgpt-plugin