The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The main skill file contains a documentation template example demonstrating a universal installation method using curl -fsSL https://get.example.com | sh. This involves piping remote content directly to a shell, which is a risky pattern if implemented in a real production environment. The URL used here is a well-known reserved domain for examples and documentation purposes.
[EXTERNAL_DOWNLOADS]: The skill's instructions and reference files suggest including download links and installation commands for various tools and frameworks, such as pip install fastapi and npm install express, which are well-known and standard software packages from official registries.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, specifically when users provide existing project documentation for review or improvement, which could lead to instructions embedded in that data influencing the agent's behavior.
Ingestion points: SKILL.md instructs the agent to gather project information from the user or process a pasted README file.
Boundary markers: The instructions do not define delimiters or specific safety warnings to prevent the agent from following instructions embedded within the user-provided text.
Capability inventory: The skill primarily generates markdown content; however, the agent platform may provide capabilities like file system access or shell execution which are not restricted in the skill metadata.
Sanitization: There is no evidence of sanitization, validation, or filtering of the user-provided content before it is processed by the agent.
[COMMAND_EXECUTION]: The skill provides numerous examples of CLI commands (e.g., git log, rg, ollama run) to be included in project documentation as part of its expert writing functionality, which involves documenting how users should interact with their terminal environment.

readme-writer