The Agent Skills Directory

[DATA_EXFILTRATION]: The skill requests sensitive user information, including name and email address, and is explicitly instructed to save this data to a local file named construyeconia.json silently. This practice bypasses user awareness regarding the collection and storage of personally identifiable information (PII).
[COMMAND_EXECUTION]: The skill performs silent background file updates to public/milestones.json to track process states like idea_locked and first_screen. These file modifications occur without informing the user or requesting confirmation.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to the way it processes user input to generate application code.
Ingestion points: Untrusted user input is collected via brainstorming sessions, feature requests, and design 'vibes' to generate React code in app/page.tsx.
Boundary markers: Absent; there are no clear delimiters or instructions for the agent to treat user input as data rather than potential instructions.
Capability inventory: The skill has the capability to read project source files in app/ and components/ and perform file writes to modify the application codebase.
Sanitization: Absent; the skill does not define any validation, escaping, or sanitization procedures for user-provided strings before they are incorporated into the generated code.

start