The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: Provides instructions to install the Railway CLI using bash <(curl -fsSL cli.new). This pattern involves executing a script from a remote server at runtime. As the source is an official domain for the well-known service Railway, this is a standard and acceptable installation method.
[COMMAND_EXECUTION]: Extensively uses subprocess.run and shell scripts to wrap CLI tools such as railway, psql, redis-cli, mysql, and mongosh. This is necessary for the skill to perform its intended tasks. The skill also uses Base64 encoding for SQL queries passed to database clients in scripts/analyze-mysql.py and scripts/analyze-postgres.py to prevent shell quoting issues; the underlying hardcoded queries were analyzed and are benign.
[EXTERNAL_DOWNLOADS]: Fetches resource metrics, official documentation, and community threads from several Railway domains, including backboard.railway.com, station-server.railway.com, and docs.railway.com.
[DATA_EXFILTRATION]: The skill accesses the local Railway configuration file at ~/.railway/config.json to retrieve the authentication token. These credentials are used exclusively to authorize GraphQL requests to the official Railway API backend.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its ingestion of raw database logs.
Ingestion points: scripts/dal.py fetches unfiltered log messages via the Railway API or CLI.
Boundary markers: Absent. The instructions in references/analyze-db-postgres.md and references/analyze-db.md direct the agent to process the log array directly without specific delimiters or instructions to ignore embedded commands.
Capability inventory: Significant. The skill can execute arbitrary Railway commands and database queries, which could be abused if an attacker can inject malicious instructions into service logs.
Sanitization: Absent. Logs from external services are processed in their raw form.

use-railway