The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill instructions are focused on legitimate content management tasks and adhere to the project's data architecture.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a potential attack surface as it processes existing project files to maintain content consistency across the corpus.
Ingestion points: The agent reads existing Markdown and MDX files from the src/data directory and schema definitions from src/content/config.ts using read_file and search_code tools.
Boundary markers: There are no explicit prompt delimiters for untrusted data; however, the skill includes a specific security guardrail (ASI06) that instructs the agent to treat retrieved research as untrusted and verify claims against live sources.
Capability inventory: The skill is authorized to use read_file, write_file, edit_file, create_file, and search_code to manipulate the project's content files.
Sanitization: No automated sanitization is implemented; the skill relies on the agent's internal safety guardrails and the mandatory human-in-the-loop editorial review process described in the 'Output Contracts'.

write-real-estate-astro-data