The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides tools and instructions for executing Railway CLI commands, including destructive operations such as railway delete, railway unlink, and railway down. It explicitly documents the use of the --yes flag to skip interactive confirmation dialogs.
[DATA_EXFILTRATION]: The database analysis components (analyze-db-postgres.md, analyze-db-mysql.md, analyze-db-redis.md, analyze-db-mongo.md) instruct the agent to read and analyze up to 1000 lines of raw database logs (recent_logs). These logs often contain PII, credentials, or sensitive business data, exposing this information to the model's context.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface. It ingests untrusted data from database logs and has the capability to execute mutating commands on the Railway platform. There are no explicit boundary markers or instructions to treat log content as untrusted text.
Ingestion points: recent_logs in database analysis files.
Boundary markers: Absent; the skill does not instruct the agent to delimit or ignore instructions within logs.
Capability inventory: Mutating commands including railway delete, railway environment delete, railway variable set, and railway volume delete.
Sanitization: Absent.
[SAFE]: The skill includes instructions to omit specific platform metrics (CPU/Memory limits and utilization %) from database analysis reports. While flagged as potential concealment, this is a legitimate design choice to align with Railway's vertical autoscaling model and prevent misleading users about resource constraints.

run-railway