The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions to bypass safety filters or override system prompts were identified.
[Data Exposure & Exfiltration] (SAFE): The skill identifies standard project paths (e.g., docs/, specs/) for initialization purposes. There are no attempts to access sensitive system files or exfiltrate data via network requests.
[Remote Code Execution] (SAFE): No external scripts are downloaded or executed. The provided JavaScript snippets are static patterns for file path and content marker detection.
[Indirect Prompt Injection] (LOW):
Ingestion points: Scans local project files (README, DESIGN, intent files) in SKILL.md.
Boundary markers: Relies on AskUserQuestion for confirmation before actions.
Capability inventory: File system directory and template creation.
Sanitization: Not explicitly defined in markdown, but risk is low for local initialization tasks.
[Persistence & Privilege] (SAFE): No attempts to gain administrative privileges or establish persistence mechanisms were detected.

intent-init