The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external HTML and local markdown content collections. This presents a surface for indirect prompt injection if the processed content contains malicious instructions intended to manipulate the agent's behavior during summary generation or site review.
Ingestion points: Production domain HTML, src/content collections, public/llms.txt, and public/llms-full.txt (referenced in SKILL.md and technical.md).
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings to isolate ingested content from the agent's internal logic.
Capability inventory: The skill performs file writes (generating metadata) and executes local scripts/CLI tools (SKILL.md, examples/cases.md).
Sanitization: The instructions include rules to exclude secrets, credentials, and private data, but there is no mechanism to sanitize or escape the content itself for prompt security.
[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands and Node.js scripts for validation and file generation purposes, specifically astrofy check --category geo and node scripts/generate-llms.mjs. While these are intended behaviors for site maintenance, they involve local command execution.

astrofy-geo