Agent skills
Agent skills drive gplay from natural-language prompts. Each skill is a
folder with a SKILL.md that documents the intent, the gplay commands it
runs, and the safety rails it enforces. They live in a companion repository:
PollyGlot/google-play-cli-skills.
Install
Section titled “Install”npx skills add PollyGlot/google-play-cli-skillsWorks with Claude Code and other agent frameworks that read the skills
format.
The roster
Section titled “The roster”One skill per shipped namespace, plus a foundation skill for cross-cutting conventions:
| Skill | Drives |
|---|---|
gplay-cli-usage | Cross-cutting conventions: output, exit codes, auth resolution (foundation) |
gplay-setup | Auth onboarding — service account, login, doctor |
gplay-apps | Apps registry + app details |
gplay-release-flow | upload / promote / staged rollouts |
gplay-tracks | Tracks + closed-track testers |
gplay-reviews | Review triage and replies |
gplay-metadata-sync | Store listings + images sync |
gplay-compliance | Data Safety declarations |
gplay-team | Users, grants, permission vocabulary |
Two more are gated until their CLI surfaces ship: gplay-vitals
(#49) and
gplay-subscription-management
(#51).
Why skills instead of raw prompting?
Section titled “Why skills instead of raw prompting?”A skill encodes the safe path for a workflow: which command sequence,
which rehearsal steps (--dry-run first), which acknowledgment flags exist
and when a human should approve them. The
agent contract explains the CLI-level
primitives the skills build on.