The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install the pgbot utility by downloading a script from https://pgbot.dev/install and piping it directly into the shell (sh). It also suggests running the tool via npx, which involves executing remote code from the npm registry.
[EXTERNAL_DOWNLOADS]: The skill fetches resources from external locations during the setup phase, specifically from the pgbot.dev domain, the pgrundev/tap Homebrew repository, and the npm package registry.
[COMMAND_EXECUTION]: The skill is designed to execute the pgbot CLI tool with various subcommands (inspect, queries, tables, vacuum, indexes, tune) to analyze database performance and health.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a PostgreSQL database, such as query strings and schema information, which represents an indirect injection surface.
Ingestion points: Database query logs, table names, and index information are read via pgbot and processed by the agent in SKILL.md.
Boundary markers: The skill lacks explicit instructions or delimiters to isolate processed database data from the agent's core instructions.
Capability inventory: The skill can execute shell commands through the pgbot CLI and interpret the resulting data to generate database management recommendations.
Sanitization: The tool claims to perform PII-free normalization of query text, reducing but not eliminating the risk of embedded instructions.

postgres-diagnostics