The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon untrusted project data including package.json, requirements.txt, Cargo.toml, and go.mod to identify tech stacks and generate Dockerfiles.
Ingestion Points: SKILL.md (Step 1 under Setting Up a New Project) and references/dockerfile-recipes.md instruct the agent to read dependency files from the user's repository.
Capability Inventory: The skill uses Bash to execute railway up, railway link, and railway variables set. It also uses Write and Edit to generate Dockerfiles and railway.toml configurations.
Boundary Markers: No explicit delimiters or instructions are provided to the agent to ignore potentially malicious content embedded in these configuration files (e.g., a malicious package name designed to trigger specific agent behavior).
Sanitization: There is no explicit sanitization step mentioned for the data read from these external files before using them to generate build scripts or CLI commands.

railway-deploy