The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external content by fetching documentation from URLs defined in the local AGENTS.md file. This creates an attack surface where a malicious project could redirect the agent to fetch instructions from an attacker-controlled source.
Ingestion points: The skill reads potential documentation addresses from the local AGENTS.md file (referenced in the 'The docs' section of SKILL.md).
Boundary markers: No specific boundary markers or instructions to ignore instructions within the fetched content are provided.
Capability inventory: The agent is authorized to install packages (npm install), modify project files (AGENTS.md, React screens), and make network requests (curl).
Sanitization: There is no evidence of validation or sanitization performed on the fetched Markdown content before processing.
[PROMPT_INJECTION]: The instructions explicitly direct the agent to use curl to bypass the constraints of standard web fetch tools, specifically noting that curl will allow access to localhost which other tools might refuse for security reasons.
[EXTERNAL_DOWNLOADS]: The skill retrieves the @tryghost/shade package from the npm registry and fetches documentation files from the vendor-owned domain shade.ghost.org.
[COMMAND_EXECUTION]: The skill utilizes shell commands for package management (npm install) and for retrieving remote documentation resources (curl).

build-ghost-app-ui