Stats
Actions
Tags
From persona-forge
Build just the evidence-grounded persona from your Claude Code history (no skill install).
How this command is triggered — by the user, by Claude, or both
Slash command
/persona-forge:persona [--days N] [--name "Your Name"]This command is limited to the following tools:
The summary Claude sees in its command listing — used to decide when to auto-load this command
Build a persona of this user from their own Claude Code transcripts — the analysis half of **persona-forge**, without generating or installing a skill. 1. **Extract** (local, no network): Read the summary, `stats.json`, and `manifest.json`. If zero sessions, stop and say so. 2. **Analyze** — spawn parallel `Task` subagents over the `corpus__*.txt` files (one per large file, group small ones, cap ~8). Each returns: identity, domains, tech stack, workflow patterns, communication style (with verbatim examples), do's, don'ts/corrections, engagement signals (flow vs. friction), recurrin...
Build a persona of this user from their own Claude Code transcripts — the analysis half of persona-forge, without generating or installing a skill.
Extract (local, no network):
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/extract_corpus.py" --out ~/.claude/persona-forge/latest $ARGUMENTS
Read the summary, stats.json, and manifest.json. If zero sessions, stop and say so.
Analyze — spawn parallel Task subagents over the corpus__*.txt files (one per large file, group small ones, cap ~8). Each returns: identity, domains, tech stack, workflow patterns, communication style (with verbatim examples), do's, don'ts/corrections, engagement signals (flow vs. friction), recurring tasks, standout quotes. Ground everything in the text; preserve their voice/typos in quotes.
Synthesize — merge into ONE persona, cross-checked against the hard numbers in stats.json (median/avg length, interrupts, peak hours, top projects, slash habits). Include an engagement model derived from their data: what keeps them engaged, what makes them disengage, when an agent should stop/ask vs. act, and the response shape they prefer.
Write ~/.claude/persona-forge/latest/persona.md and show the user a tight summary (identity, signature traits, engagement model, tech profile, a few real quotes).
End by offering: /persona-forge:personalize to turn this into an installed, self-updating skill.
npx claudepluginhub akhatua2/persona-forge --plugin persona-forge/personaCreates a detailed buyer persona for a product or segment through interactive parameter collection, then outputs a structured persona profile.