The Agent Skills Directory

[PROMPT_INJECTION]: The methodology requires the agent to read and process multiple documentation files (e.g., docs/product/journeys.yaml, docs/feature/*/discuss/user-stories.md) to generate executable code. This pattern of using external documentation as a source for code generation is a known surface for indirect prompt injection. The agent has file system write capabilities and can execute local commands like 'nwave-ai' and 'pytest', though no explicit boundary markers or sanitization procedures for the ingested documentation are described.
[COMMAND_EXECUTION]: The skill facilitates the use of the 'nwave-ai' CLI and common test runners such as 'pytest' and 'vitest'. These commands are executed locally within the user's project environment to manage test scaffolds and verify implementation status, consistent with the skill's purpose.
[DATA_EXFILTRATION]: The skill instructs the agent to read local configuration files, specifically '~/.nwave/global-config.json'. This access is used to determine project settings like documentation density modes and is an expected interaction within the tool's integrated development ecosystem.

nw-distill