The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill automates the installation of the Ravion CLI by downloading a script from a remote GitHub repository and piping it directly to the shell (sh). This occurs in SKILL.md and setup.md using the command curl -fsSL https://github.com/ravionhq/cli/releases/latest/download/install.sh | sh.- [EXTERNAL_DOWNLOADS]: The skill installs a Model Context Protocol (MCP) server from an external URL using npx. Specifically, in setup.md, it uses npx add-mcp https://www.ravion.com/docs/mcp --name ravion-docs to modify the agent's execution environment.- [CREDENTIALS_UNSAFE]: The skill provides instructions to the agent to perform signups by passing email and password credentials as plain-text command-line arguments. This pattern (ravion signup --email <email> --password <password>) is found in SKILL.md and setup.md and exposes sensitive user credentials to shell history, process listings, and system logs.- [PRIVILEGE_ESCALATION]: The skill automates the creation of cross-account IAM roles in AWS with elevated privileges. It uses the aws cloudformation create-stack command with the CAPABILITY_NAMED_IAM flag, as seen in setup.md, which allows for the creation of IAM resources that can grant significant permissions.- [COMMAND_EXECUTION]: The skill uses broad shell command execution capabilities to drive infrastructure provisioning, including aws, ravion, and npx commands that perform high-impact actions in the cloud environment.- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it parses manifest files from external repositories to drive deployment logic.
Ingestion points: Manifest files such as package.json, pyproject.toml, go.mod, Gemfile, and Dockerfile in the project root as described in SKILL.md.
Boundary markers: There are no delimiters or specific "ignore embedded instructions" warnings for the agent when processing these external files.
Capability inventory: The skill possesses extensive capabilities, including system shell access and infrastructure management tools (aws, ravion, curl, npx).
Sanitization: No sanitization or strict schema validation is described for the content extracted from manifest files before it influences command generation.

use-ravion