The Agent Skills Directory

[PROMPT_INJECTION]: The skill describes a method for hiding agent-created files from the user's git status output by modifying .git/info/exclude. This is a concealment pattern that can be used to hide modifications or new files from user oversight, even if intended for legitimate environment isolation.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to gather setup requirements by reading various repository files which may contain untrusted content. This creates a vulnerability where malicious data in a repository could influence the shell scripts generated by the agent.
Ingestion points: The agent is directed to read package.json, pnpm-lock.yaml, Cargo.toml, go.mod, requirements.txt, pyproject.toml, Gemfile, README.md, CONTRIBUTING.md, AGENTS.md, and CI workflow configurations.
Boundary markers: No specific delimiters or instructions to ignore embedded prompts within the analyzed files are provided.
Capability inventory: The skill involves creating and executing bash scripts, installing system packages via apt-get, and modifying the user's shell profile (~/.bash_profile).
Sanitization: The instructions do not include steps to sanitize or validate the content extracted from repository files before using it to construct shell commands.
[COMMAND_EXECUTION]: The skill's primary function involves generating and running shell scripts for environment configuration, tool installation, and service management.
[EXTERNAL_DOWNLOADS]: The skill references documentation and configuration resources located on the author's official domain (ampcode.com).

orb-setup