Updates agent identity files through an interactive interview backed by a shared answer bank.
npx skills add https://github.com/AMC-JTC/gbrain-1 --skill soul-audit-amc-jtc
Please help me install this Agent Skill. Skill: soul-audit Source: https://github.com/AMC-JTC/gbrain-1/tree/main/plugin/skills/soul-audit Command: npx skills add https://github.com/AMC-JTC/gbrain-1 --skill soul-audit-amc-jtc
gbrain bootstrap interview --set/--skip into state/interview.json, the same store used by first-boot bootstrap. - Read-back confirmation gate: Requires showing answers to the user and confirming a hash before any render, preventing unreviewed changes. - Safe re-rendering: Re-renders SOUL.md, USER.md, ACCESS_POLICY.md, HEARTBEAT.md, and AGENTS.md with --force, backing up old files under .gbrain-bootstrap-backups/ first. - Six independent phases: Identity, vibe calibration, mission mapping, user profile, access boundaries, and operational cadence — each re-runnable on its own. - Use Case: A user wants to change their agent's communication style from formal to direct. The skill updates VOICE_REGISTER in the answer bank, confirms the change with the user, and re-renders only SOUL.md with a backup of the previous version. ## Quick Start Ask the agent to run a soul audit to update its identity, then answer the interview questions for the phase you want to change.High-intent search queries and answers about installing and using this skill.
Run the soul audit skill and complete the vibe calibration phase. It records your answers into the shared interview answer bank, confirms them with you, then re-renders SOUL.md through the bootstrap render engine.
Update the underlying answer in the interview bank with gbrain bootstrap interview --set, then re-render with gbrain bootstrap render --only SOUL.md --force. Hand-edits are overwritten by the next render, so always edit the answer, not the file.
No. Rendering with --force first backs up the existing file under .gbrain-bootstrap-backups/<timestamp>/ before writing the new version, so every change is recoverable.
Yes. Each of the six phases is independent and re-runnable. Updating a single answer and re-rendering only the file it feeds is a complete, valid run.
The render gate requires a read-back ritual: the full answer set must be shown to the user and confirmed with the hash printed by --status. Confirming a hash for answers the user never saw fails by design.
garrytanUpdates agent identity files through an interactive interview backed by a shared answer bank.
x-cmdDelegate tasks to the heartbeat agent by editing HEARTBEAT.md.
xpert-aiDeploys prototypes to Sites or Vercel and generates shareable URLs.
haah-ingOrganize personal contacts into a local searchable dataset of people and organizations.