The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to read the user's shell configuration file (~/.zshrc) to extract the DEEPAPI_READONLY_DB_URL. This file is a sensitive system location that often contains various secrets, API keys, and environment variables.
[COMMAND_EXECUTION]: The skill uses shell commands, including rg (ripgrep) to parse files and psql to interact with a production database. The database connection string is dynamically constructed from environment variables.
[DATA_EXFILTRATION]: The skill is designed to access and process live production data from Supabase Postgres. This includes tables like workspaces, service_requests, and dashboard_workspace_members which contain sensitive information, including email addresses and usage logs.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from a production database.
Ingestion points: SQL query results from public.service_requests and other tables in SKILL.md.
Boundary markers: Absent; there are no instructions to delimit or treat database output as untrusted content.
Capability inventory: Shell command execution via psql and rg is available in SKILL.md.
Sanitization: Absent; the skill does not specify any filtering or escaping of database content before it is processed by the agent.

read-prod-database