The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external information such as "agreed requirements," "technical decisions," and the content of the "existing repository" to generate or update agent instructions (e.g., in AGENTS.md). This creates a surface for indirect prompt injection attacks.
Ingestion points: Processes external technical and product documentation provided by the user or found within the repository (e.g., docs/context-pack.md).
Boundary markers: The instructions do not specify the use of delimiters or specific warnings to ignore instructions embedded within the processed data.
Capability inventory: The skill has access to powerful tools including Bash, Write, Edit, and Grep, which allow it to modify files and execute shell commands.
Sanitization: There is no mention of input sanitization or validation for the content being processed into instructions.
[EXTERNAL_DOWNLOADS]: The skill utilizes an external command-line tool which may result in remote code download and execution.
Evidence: The instructions call for running npx vibeworkflow --dry-run --json and npx vibeworkflow doctor.
Risk: The use of npx automatically downloads the package from the NPM registry if it is not present in the local environment, executing code from an external source.
[COMMAND_EXECUTION]: The skill performs command execution through the Bash tool to interact with the environment and the external CLI tool.
Evidence: Executing the vibeworkflow CLI via npx to inspect configuration and initialize setup.

vibe-agents