The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents the use of 'Live Content Collections' (in references/new-features.md), which allows the agent to ingest and process data from external API endpoints (e.g., https://api.example.com/news). This capability introduces a standard indirect prompt injection surface where external content could potentially influence the agent's context.
Ingestion points: Content loaders defined in src/content.config.ts fetching from external URLs.
Boundary markers: The skill does not include instructions for using delimiters to isolate external data.
Capability inventory: The agent is instructed to use file system access and shell commands (npm, astro) for project management.
Sanitization: No data validation or sanitization methods are specified for the external content.
[EXTERNAL_DOWNLOADS]: The documentation references official Astro tools for initialization and maintenance, such as npm create astro@latest and npx @astrojs/upgrade, which are standard for the framework and originate from a trusted project source.
[COMMAND_EXECUTION]: The skill includes instructions for running common development and build commands, such as astro dev and astro build, which are intended for standard local environment usage.
[METADATA_POISONING]: Several reference files (e.g., references/installation.md, references/routing.md) inconsistently refer to 'Astro 6' despite the skill being focused on 'Astro 7'. Additionally, the mention of a non-standard src/fetch.ts reserved file and fetchFile configuration option in references/routing.md could lead to project configuration errors or hallucinations if these features do not exist in the target framework version.

astro-7