Analysis updated 2026-05-18
Ask an AI agent to promise it will never touch a database schema during a session
Set a strong level of promise for a rule you really care about, like avoiding hidden vulnerabilities
Undo a previous promise using a matching unlock phrase the agent agrees to
Add the same rule enforcement idea to any AI agent that supports skill files
| alesimula/pinkypromise | 00kaku/gallery-slider-block | 04amanrajj/netwatch | |
|---|---|---|---|
| Stars | 0 | — | 0 |
| Language | — | JavaScript | Rust |
| Last pushed | — | 2021-05-19 | — |
| Maintenance | — | Dormant | — |
| Setup difficulty | easy | easy | moderate |
| Complexity | 1/5 | 2/5 | 3/5 |
| Audience | developer | general | ops devops |
Figures from each repo's GitHub metadata at analysis time.
Installs as a Claude Code plugin marketplace, or as a single skill file for other agents.
pinkypromise is a playful add-on for AI coding agents that turns a rule into something closer to a spoken promise than a written policy. Instead of writing a rule into a config file, you ask your agent to pinky swear that it will follow it, such as never touching a database schema during a session. The agent can accept or decline the swear, but once it agrees, the idea is that it should stick to that rule for the rest of the session without quietly backing away from it. The tool defines a few levels of how seriously the promise is meant to be taken, from a light gentleman's agreement up to a stronger version meant for rules the user really cares about. A swear can only be undone through a matching unlock phrase, and even then the agent has to agree to dissolve it, so ending a promise is treated the same way as making one, as a two sided exchange rather than a simple command. The README frames the whole idea with humor, describing the pinky swear in playful, exaggerated language, while also including a more serious sounding passage that calls it a binding security commitment. Readers should treat that framing as part of the joke rather than an actual technical guarantee, since nothing in the project changes how an AI model is restricted at a lower level. Installing it for Claude Code works through the plugin and marketplace system built into the tool, since the repository ships its own marketplace file. For other AI agents, the entire behavior lives in a single skill file that can be added by hand however that agent loads skills. The project is released with permissive, informally worded terms that amount to letting anyone use, modify, or share it freely.
A playful plugin that has an AI coding agent pinky swear to follow a rule for the rest of a session, with humorous levels of seriousness.
Use, modify, and share this freely without restriction.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.