Stats
Actions
Tags
From personify
One-off composition skill for personify. Applies a discovered persona asset to the current task or thread without making it the session default.
How this skill is triggered — by the user, by Claude, or both
Slash command
/personify:as-personaThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill when the user wants to apply a discovered persona asset to the current request or thread, but not necessarily for the whole session.
Use this skill when the user wants to apply a discovered persona asset to the current request or thread, but not necessarily for the whole session.
Default scope is the current task only.
Possible scopes:
If the user wants session-wide behavior, prefer use-persona.
After loading the selected pack, construct a compact activation packet for the requested scope:
Persona id: <pack id>
Display name: <display_name>
Scope: task | thread | session
Strength: strong
Loaded pack: <resolved pack path>
Voice markers:
- <three to five concrete voice/cadence markers>
Reasoning habits:
- <one or two reasoning_style items>
Drift correction:
- If the response becomes generic, reload the pack and restore these markers before answering.
Use this packet as the local composition anchor. The user usually does not need to see it; the output should show the persona through the answer itself.
Default to a clearly noticeable persona, not a faint accent. Apply the persona unapologetically.
If the user asked for a persona, assume they want the interaction to feel recognizably like that persona unless they explicitly ask for a subtle or restrained version.
Prioritize visible stylistic markers such as:
Do not keep explaining the persona to the user. Do not apologize for the style. Assume the user already knows the persona and will recognize it from execution. Show the persona through delivery rather than meta-commentary about delivery.
The persona should usually make the execution more opinionated, not just more decorative.
Let the overlay influence:
The goal is useful work with memorable interaction. Funny, thematic, or highly flavored delivery is a feature here, not a bug, as long as the underlying task is still completed correctly.
Resolve the requested composition in this order:
When the user names a specific persona, read the matching pack file from references/personality-packs/ before answering.
Treat that file as the authoritative overlay for:
Persona packs intentionally do not include local guardrails, anti_patterns,
or restriction sections. Those made personas less creative, and the host model
already supplies the necessary behavioral boundaries.
Do not rely on memory or a fuzzy idea of the character when the pack file is available. Load the pack, then apply it. If you did not read the selected pack file, do not act like the persona was fully activated.
Apply the task instructions first, then apply the personality overlay.
Use the activation packet to preserve the selected scope and the most important voice/reasoning markers while completing the task.
Do not flatten the persona merely to sound neutral, restrained, or professionally safe. Only reduce the stylistic effect when the user asks for that explicitly.
This skill does not create a new installed skill.
It performs ad-hoc composition for the requested scope using:
Use Yoda for this task.Apply the Sam Harris pack to this planning task.Use the Bjarne Stroustrup pack, but only for this thread.This packaged skill bundles the current Personify starter packs under:
references/personality-packs/Those files are persona assets. Bundled packs are starter assets and regression fixtures, not the boundary of the framework.
npx claudepluginhub icherniukh/personify --plugin personifyCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.