The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from community sources through scripts/search-block-party.js and scripts/get-block-structure.js. These sources contain user-contributed titles, descriptions, and HTML structures which could potentially contain malicious instructions intended to influence the agent's behavior. Mandatory Evidence Chain: (1) Ingestion Points: The skill fetches the community Block Party index and specific block documentation pages from aem.live. (2) Boundary Markers: SKILL.md contains a specific 'External Content Safety' section instructing the agent to treat all fetched content as untrusted and ignore embedded directives. (3) Capability Inventory: The skill provides search and structural parsing capabilities but lacks dangerous functions like eval() or file-write operations on the external data. (4) Sanitization: The search script explicitly filters for entries marked as 'approved' in the community index.
[EXTERNAL_DOWNLOADS]: The skill fetches configuration and data from github.com/adobe and aem.live. These are official Adobe/AEM Edge Delivery Services domains and are recognized as trusted vendor resources for this skill's context. The skill also includes dependencies on future-dated package versions (e.g., jsdom v30.1.0) in its package-lock.json, which is consistent with the skill's own 2026 timestamp in the CHANGELOG.md and does not represent a current security risk.

block-collection-and-party