The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to ingest and analyze untrusted source code and data hooks from the project to generate 'State Machine' workbenches.
Ingestion points: The agent reads component code, data hooks, and project configurations directly from the local file system (references/states/state-machine.md).
Boundary markers: The instructions do not provide specific boundary delimiters or safety prompts to prevent the agent from following instructions hidden in comments or metadata within the ingested source code.
Capability inventory: The skill empowers the agent to write new files (scratch routes, fixtures) and perform visual edits to existing application code (CSS, TSX).
Sanitization: No explicit sanitization or filtering logic is prescribed for the data ingested from the project code, although the skill does recommend isolating mock data from live production data.
[EXTERNAL_DOWNLOADS]: The SOURCE-MANIFEST.json file references several external GitHub repositories (specifically jakubkrehel, emilkowalski, and backnotprop accounts) as upstream sources for bundled content.
Context: These URLs are documented as part of the skill's build-time manifest for content patching and provenance, rather than being used for runtime script execution.

atlas-craft