The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool to execute gh release view for its version checking mechanism. This command targets the vendor's repository on GitHub to identify if a newer version is available.
[EXTERNAL_DOWNLOADS]: During the activation flow, the skill retrieves metadata from github.com/tonylofgren/aurora-smart-home. This is a vendor-owned resource on a well-known service (GitHub) used for legitimate maintenance checks.
[PROMPT_INJECTION]: The skill architecture presents an indirect prompt injection surface by ingesting untrusted user input to drive a complex multi-agent workflow.
Ingestion points: User requirements are collected in SKILL.md (Step 1) and stored in the aurora-project.json snapshot file, which serves as the source of truth for all specialist agents.
Boundary markers: The system uses structured JSON snapshots and specialist personae to maintain context boundaries, although it lacks specific instructions to ignore malicious directives embedded within the requirements text.
Capability inventory: The specialist sub-agents (e.g., souls/ada.md, souls/volt.md, souls/sage.md) possess significant capabilities including file creation (Write), file modification (Edit), and shell execution (Bash).
Sanitization: The skill includes an extensive suite of validator specifications (e.g., secrets-validator.md, async-correctness-validator.md) that specialists are required to run against their output before delivery to identify and mitigate security risks such as hardcoded credentials.

aurora