The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands (ln -s on Linux/macOS or New-Item on Windows) to create symbolic links between documentation files. It specifically requests the use of "PowerShell 管理员模式" (Administrator mode) on Windows to perform these operations.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads untrusted content from various project configuration files to generate instructions for other AI agents.
Ingestion points: The skill scans package.json, pyproject.toml, go.mod, Cargo.toml, composer.json, Makefile, and Dockerfile to identify the tech stack and project structure.
Boundary markers: The provided AGENTS.md template does not utilize specific delimiters or escape sequences to isolate data extracted from configuration files from the rest of the instruction set.
Capability inventory: The skill has the capability to read local files, write new documentation files, and execute shell commands to create symbolic links.
Sanitization: There is no mention of sanitizing or validating the content retrieved from project configuration files before it is incorporated into the generated AI instructions.

init-agent