package-postgres-agy-green
Warn
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
greenscript dynamically downloads code from the author's repositories at runtime, specifically fromhttps://github.com/getcolors/postgres-agy.git,https://github.com/getcolors/green.git, andhttps://github.com/getcolors/once.git. - [REMOTE_CODE_EXECUTION]: The script uses
babashka.deps/add-depsto fetch code from remote Git repositories and then executes it usingrequiring-resolve, allowing for execution of code not present in the local skill files. - [COMMAND_EXECUTION]: The skill dispatches administrative commands to a PostgreSQL cluster over SSH using aliases defined in the local
~/.ssh/configfile. - [DYNAMIC_EXECUTION]: The entry point script is a launcher that dynamically loads its core functionality and dependencies at runtime based on hardcoded Git SHAs or local environment variable overrides.
- [INDIRECT_PROMPT_INJECTION]: 1. Ingestion points: The skill reads project configuration from
colors.ymland processes secrets fromCOLORS_PAR_*environment variables. 2. Boundary markers: Not present in the launcher script; delimiters are presumably handled by the remote library. 3. Capability inventory: The skill has network access and shell execution capabilities via Babashka and SSH. 4. Sanitization: Sanitization is not visible in the launcher and depends on the remote dependency's implementation.
Audit Metadata