Add Alter — digital persona skill (local-first, self-benchmarking)

[JasonQuantum]

What this adds

One row in Community Skills → Individual Skills for Alter (Apache-2.0).

Alter builds a digital persona of the user from a chat interview (voice memos transcribed locally via whisper.cpp), their documents, and their exported AI chats — then chats and drafts messages/emails in their voice. It keeps improving from corrections; when new information contradicts stored facts it asks instead of overwriting. Eight sealed validation questions are held out of memory so the persona can benchmark itself against the real person and report the gaps.

Social proof / signals

• Live on skills.sh: https://skills.sh/alter-persona/alter (npx skills add alter-persona/alter — the pack follows the agentskills.io layout, skills/alter/SKILL.md, validated with skills-ref validate)
• Cross-platform: same SKILL.md loads as ready on OpenClaw and installs into Claude Code via the skills CLI symlink; also runs on Hermes (hermes skills tap add alter-persona/alter)
• Released v0.1.1 with protected release tags; full local-first setup documented (Docker Postgres + pgvector, whisper.cpp, Ollama; ElevenLabs optional for voice replies)
• Young project (released this month) — happy to have it re-categorized or held to whatever bar you prefer.

🤖 Generated with Claude Code