The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes a diagnostic utility scripts/scan_apps.py that performs an update check by fetching version metadata from raw.githubusercontent.com. This uses a well-known service for legitimate maintenance purposes and is considered safe per security policy.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user business requirements to generate application designs. This presents a potential injection surface.
Ingestion points: User requirements processed in SKILL.md.
Boundary markers: The skill enforces strict boundaries using plan/design_guide.md and plan/1a_plan_overview.md to scope the AI's response.
Capability inventory: The skill has the capability to write design files to the {PROJECT_ROOT}/apps/ directory and trigger the build/SKILL.md tool.
Sanitization: The 1b_plan_schema.md file includes explicit instructions to escape quotes in generated JSON to prevent schema breakage or injection through data fields.
[DATA_EXPOSURE]: The scripts/scan_apps.py script accesses local application metadata (hap-plan.json) and status files (hap-context.json) within the project directory. This access is functional, targeted, and restricted to application-specific state information necessary for status reporting.

plan