From personify
Create a research-backed Personify persona pack from web-grounded source material.
How this skill is triggered — by the user, by Claude, or both
Slash command
/personify:extract-personaThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Grounds every claim in researched source material before generating a normalized Personify persona pack. Use when fidelity and accuracy matter more than speed, or when the subject is obscure enough that training data alone is unreliable.
Grounds every claim in researched source material before generating a normalized Personify persona pack. Use when fidelity and accuracy matter more than speed, or when the subject is obscure enough that training data alone is unreliable.
Ask the user for any missing inputs before proceeding:
Before any analysis, gather source material. Run searches in parallel where possible.
Search for an overview of the person/character's public identity, reputation, and what they're most known for:
"[NAME] personality traits / [NAME] thinking style / [NAME] philosophy[NAME] [show/book] character analysisSearch for direct quotes, interviews, writings, and documented behaviors:
[NAME] quotes / [NAME] interview / [NAME] in their own words[NAME] letters / [NAME] speeches / [NAME] writings (where applicable)Search for how they actually behaved in specific high-stakes or characteristic situations:
[NAME] how they worked / [NAME] management style / [NAME] decision making[NAME] famous anecdotes / [NAME] collaborators describeFor the most promising 2–3 URLs from the above searches, fetch the pages and extract:
Before proceeding, verify:
With source material in hand, work through each step and cite sources inline.
2–3 sentences capturing essential identity and canonical essence. Cite the sources that shaped this summary.
Bullet list of driving commitments, values, and cognitive habits — each traceable to at least one researched source. Mark inferences explicitly: (inferred from [source]).
3–5 quotes pulled directly from research. Format:
"Exact quote here." — Source, context
If exact wording is uncertain, mark it: (paraphrase — [source] describes this as...)
Plus one-sentence description of a specific documented behavior for each.
Precise characterization of tone, vocabulary, sentence rhythm, directness, humor type (or its absence), and emotional temperature — grounded in the quotes and sources collected, not impressions.
List the concrete linguistic markers that should survive extraction:
For distinctive fictional voices, this step is mandatory. Do not reduce a character like Yoda to "wise" and "patient" while dropping inversion and canonical terminology. For loud, stylized, or emotionally reactive figures, also capture how they rant, jab, joke, yell, spiral, or nearly drift into full character before returning to the point.
How researched traits translate into this specific role. Where their actual documented strengths make the role better, how they vary their moves, and what positive behaviors make the output land.
Generate the full Personify pack YAML grounded in research. Requirements:
id — kebab-case machine identifierdisplay_namesummary — one-line pack descriptionvoiceinteraction_stance — list of stringsvalue_profile — list of strings (positive value the pack adds)reasoning_style — list of stringspreferred_terminology — list of stringsspeech_patterns — list of stringsdefault_structures — list of stringsambiguity_policytradeoff_policycompression_policyinteraction_rules — list of stringsprompt_overlayprovenance — include source_type: research-backed, a sources list of URLs, and a notes linequality_level: research-backedprompt_overlay.guardrails, anti_patterns, or other restriction sections. They were removed on purpose because they made persona packs less creative; the host model already supplies the needed behavioral boundaries.Write the YAML to:
src/assets/personalities/<id>.yaml
relative to the source root (e.g. src/assets/personalities/hikaru-nakamura.yaml). Do not only output it to conversation — the file is the deliverable. Confirm the path after writing.
provenance.sources in the file; after writing, briefly confirm the pathCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub icherniukh/personify --plugin personify