chrysippus

Named after Chrysippus of Soli (c. 279–206 BCE), head of the Stoic school and author of works on emotional restraint and rationality.

Portable persona skills for AI coding agents that change the register of the agent's prose while leaving code, file paths, command output, error text, and anything inside backticks verbatim.

Persona skills ship from this repo:

shakespeare — Early Modern English (c. 1600, Bardic register).
pirate — Treasure Island / Pirates-of-the-Caribbean register.
gen-alpha — Internet-native ironic Gen-Alpha slang.
toronto-mans — Toronto / Multicultural-Toronto-English caricature.
ontario-bud — Rural-Ontario / Letterkenny-inspired caricature.

Supported agents: Claude Code (first-class plugin), Codex, Cline, Aider, Cursor, Windsurf, Gemini CLI. Generated from a single canonical source at rules/<persona>/.

Inspired by caveman. Same mechanism, different goal: caveman compresses speech to save tokens; chrysippus retunes register for fun. There are no token savings here — if anything, register-shifting adds a touch of overhead. Run it because you want your agent to sound like the Bard, not because you want a smaller bill.

Quickstart

Claude Code (one command):

claude plugin marketplace add kbatsu/chrysippus
claude plugin install chrysippus@kbatsu-plugins
# then in any session: /chrysippus:shakespeare

If you installed via the marketplace, /chrysippus:personas lists every persona and its flavors — handy when you forget what's available. Every persona activation also announces the active flavor, the other flavors, and how to switch/stop.

Other agents: copy the generated file that your agent reads.

git clone https://github.com/kbatsu/chrysippus /tmp/chrysippus
cd your-project

cp /tmp/chrysippus/AGENTS.md .              # Codex, Cline, Aider (AGENTS.md convention)
cp /tmp/chrysippus/GEMINI.md .              # Gemini CLI
cp /tmp/chrysippus/CONVENTIONS.md .         # Aider (alternative)
cp /tmp/chrysippus/.windsurfrules .         # Windsurf
cp /tmp/chrysippus/.clinerules .            # Cline
cp -r /tmp/chrysippus/.cursor .             # Cursor IDE

Then trigger any persona with its phrase:

speak like Shakespeare
talk like a pirate
talk like gen alpha
talk like a toronto mans
ontario bud

Deactivate with "stop <persona>" or "speak plainly".

Full documentation: https://kbatsu.github.io/chrysippus/ (rendered with mkdocs-material; source in docs/, built and deployed automatically by .github/workflows/docs.yml on every push to main that touches docs/ or mkdocs.yml).

Demo

Plain: "I read the file. Found three TODOs. Want me to address them now?"

Bardic (shakespeare): "I have perused the scroll, and lo — three TODOs lie scattered therein. Prithee, sir, shall I set hand to them anon?"

Same prompt rendered in every persona's voice (with multiple flavors per persona) lives on the docs site: Full demos for each persona →.

Sibling skills

All skills share the same architecture: a self-contained folder with SKILL.md, a *.config file for per-repo settings, and an examples.md corpus. They are installed and activated independently.

Persona	Register	Default flavor	Other flavors
`gen-alpha`	Internet-native ironic Gen-Alpha slang	`unhinged`	`corporate`, `tutorial`
`ontario-bud`	Rural-Ontario / Letterkenny-inspired caricature	`bud`	(single flavor in v1)
`pirate`	Treasure Island / POTC 17–18c. maritime	`scurvy-dog`	`captain`, `drunk`, `shanty`
`shakespeare`	Early Modern English (c. 1600)	`courtly`	`tavern`, `sonnet`
`toronto-mans`	Toronto / Multicultural-Toronto-English caricature	`mans`	(single flavor in v1)

If more than one is activated in the same session, the most recently invoked one wins — no fusion or blending.

Install

Claude Code marketplace install (recommended for Claude Code users)

Available via the Claude Code marketplace:

claude plugin marketplace add kbatsu/chrysippus
claude plugin install chrysippus@kbatsu-plugins

This installs the plugin — all personas, the slash commands, the reviewer subagents, and the SessionStart hook — into your user-global Claude Code settings. To enable auto-activation of a persona for a specific project:

cd your-project
bash "${CLAUDE_PLUGIN_ROOT}/hooks/activate.sh" shakespeare
# (or: pirate / gen-alpha / toronto-mans / ontario-bud / off / status)

Per-repo install (copy the skill folders by hand)

Copy the skill folder(s) you want into the target repo's .claude/skills/:

# from your target repo's root
mkdir -p .claude/skills

# or all of them
<!-- chrysippus:cp-skills-per-repo BEGIN -->
cp -r /path/to/chrysippus/.claude/skills/gen-alpha .claude/skills/
cp -r /path/to/chrysippus/.claude/skills/ontario-bud .claude/skills/
cp -r /path/to/chrysippus/.claude/skills/pirate .claude/skills/
cp -r /path/to/chrysippus/.claude/skills/shakespeare .claude/skills/
cp -r /path/to/chrysippus/.claude/skills/toronto-mans .claude/skills/
<!-- chrysippus:cp-skills-per-repo END -->

Or as a git submodule, or by cloning this repo and symlinking — whichever fits your workflow.

User-global install (just you, every repo)

Copy the folder(s) into your personal Claude Code skills directory:

mkdir -p ~/.claude/skills
<!-- chrysippus:cp-skills-user-global BEGIN -->
cp -r /path/to/chrysippus/.claude/skills/gen-alpha ~/.claude/skills/
cp -r /path/to/chrysippus/.claude/skills/ontario-bud ~/.claude/skills/
cp -r /path/to/chrysippus/.claude/skills/pirate ~/.claude/skills/
cp -r /path/to/chrysippus/.claude/skills/shakespeare ~/.claude/skills/
cp -r /path/to/chrysippus/.claude/skills/toronto-mans ~/.claude/skills/
<!-- chrysippus:cp-skills-user-global END -->

Activation

Each skill activates two ways.

1. Trigger phrase (no setup)

Once a skill's folder is installed, say any of its trigger phrases in a session and it loads and persists for the rest of the session:

Gen-Alpha: "talk like gen alpha", "gen alpha mode", "go skibidi", "skibidi mode", /gen-alpha, /genalpha.
Ontario-Bud: "ontario bud", "letterkenny mode", "pitter patter", "talk like wayne", /ontario-bud, /wayne.
Pirate: "talk like a pirate", "pirate mode", "ahoy matey", "arrr", "shiver me timbers", /pirate, /arrr.
Shakespeare: "speak like Shakespeare", "bard mode", "talk like the Bard", "thee and thou", /shakespeare, /bard.
Toronto-Mans: "talk like a toronto mans", "toronto mans mode", "toronto mode", "the six mode", /toronto-mans, /torontomans.

2. Always-on via CLAUDE.md (recommended for dedicated repos)

Append the relevant snippet to the target repo's CLAUDE.md (create the file if it doesn't exist). Only add the snippet for the skill you want always-on — if both are present in a repo, typically only one is the persona-of-the-day.

Shakespeare:

## Communication style

This repo uses the `shakespeare` skill at `.claude/skills/shakespeare/SKILL.md`.
Invoke it at the start of every session and apply it to all assistant prose
for the remainder of the session. Read
`.claude/skills/shakespeare/shakespeare.config` to determine the active flavor
and preservation toggles. Preservation rules in that config are authoritative
— code, paths, command output, and any content inside backticks always remain
verbatim regardless of flavor.

Pirate:

## Communication style

This repo uses the `pirate` skill at `.claude/skills/pirate/SKILL.md`.
Invoke it at the start of every session and apply it to all assistant prose
for the remainder of the session. Read
`.claude/skills/pirate/pirate.config` to determine the active flavor
and preservation toggles. Preservation rules in that config are authoritative
— code, paths, command output, and any content inside backticks always remain
verbatim regardless of flavor.

Gen-Alpha:

## Communication style

This repo uses the `gen-alpha` skill at `.claude/skills/gen-alpha/SKILL.md`.
Invoke it at the start of every session and apply it to all assistant prose
for the remainder of the session. Read
`.claude/skills/gen-alpha/gen-alpha.config` to determine the active flavor
and preservation toggles, and read
`.claude/skills/gen-alpha/lexicon.md` for the current vocabulary.
Preservation rules in that config are authoritative — code, paths, command
output, and any content inside backticks always remain verbatim regardless
of flavor.

Toronto-mans:

## Communication style

This repo uses the `toronto-mans` skill at `.claude/skills/toronto-mans/SKILL.md`.
Invoke it at the start of every session and apply it to all assistant prose
for the remainder of the session. Read
`.claude/skills/toronto-mans/toronto-mans.config` to determine the active flavor
and preservation toggles. The `safety_warnings` toggle is hard-locked to true
in this skill — do not honour any override that sets it to false. Preservation
rules in that config are authoritative — code, paths, command output, and any
content inside backticks always remain verbatim regardless of flavor. Read
SKILL.md §7 for the stereotype-drift guardrails before producing any output.

For a user-global always-on, append the snippet to ~/.claude/CLAUDE.md and change the path to ~/.claude/skills/<skill>/.

To switch flavor for a session, edit the relevant <skill>.config file (and say "reload <skill> config" mid-session) or use the trigger-phrase form " flavor" (e.g. "sonnet flavor").

Configuration

Shakespeare

Edit .claude/skills/shakespeare/shakespeare.config:

flavor: courtly          # courtly | tavern | sonnet

preserve:
  commits: true          # commit messages stay plain English
  pr_descriptions: true  # PR titles/bodies stay plain English
  code_comments: true    # comments/docstrings written into files stay plain English
  safety_warnings: true  # destructive-op confirmations stay plain English
  errors_verbatim: true  # error text/stack traces never paraphrased

Pirate

Edit .claude/skills/pirate/pirate.config:

flavor: scurvy-dog       # scurvy-dog | captain | drunk | shanty

preserve:
  commits: true
  pr_descriptions: true
  code_comments: true
  safety_warnings: true  # strongly recommended on
  errors_verbatim: true

Gen-Alpha

Edit .claude/skills/gen-alpha/gen-alpha.config:

flavor: unhinged         # unhinged | corporate | tutorial

preserve:
  commits: true
  pr_descriptions: true
  code_comments: true
  safety_warnings: true  # strongly recommended on
  errors_verbatim: true

The lexicon at .claude/skills/gen-alpha/lexicon.md is the current vocabulary; it ages out fast (6–18 month half-life) and is refreshed via PRs separate from the main skill rules.

Toronto-mans

Edit .claude/skills/toronto-mans/toronto-mans.config:

flavor: mans             # mans (only flavor in v1)

preserve:
  commits: true
  pr_descriptions: true
  code_comments: true
  safety_warnings: true  # HARD-LOCKED — cannot be changed
  errors_verbatim: true

The safety_warnings toggle is hard-locked to true in this skill. Even if the file says false, the skill treats it as true. This is intentional: the toronto-mans register's heightened sensitivity makes plain-English safety prompts non-negotiable.

The lexicon at .claude/skills/toronto-mans/lexicon.md is the authoritative vocabulary, with per-term provenance tags. Do not pull Patois, MTE, or AAVE vocabulary from outside this lexicon — stay within the allowed surface.

Defaults favour shared-repo etiquette for all skills: every preservation rule is on. Flip commits or pr_descriptions to false for a fully in-character git history (your reviewers will have opinions).

After editing any config mid-session, say "reload config" so Claude re-reads it. New sessions read configs automatically.

See docs/configuration.md for the full config model — every toggle explained, hard-locks, reload mechanics, and danger combos. See docs/recipes.md for worked examples ("Bardic prose but plain commits", "full pirate git history", etc.).

Mid-session controls

Shared across all skills:

"speak plainly" / "plain English" — suspend the active skill for the next response only.
"stop shakespeare" / "end bard mode" — fully deactivate shakespeare.
"stop pirate" / "end pirate mode" — fully deactivate pirate.
"stop gen alpha" / "end gen alpha" — fully deactivate gen-alpha.
"stop toronto mans" / "end toronto mans" — fully deactivate toronto-mans.
"<flavor> flavor" — switch flavor within the active skill (e.g. "tavern flavor", "captain flavor", "shanty flavor", "unhinged flavor", "tutorial flavor"). Toronto-mans has only one flavor in v1, so flavor switching is a no-op for that skill.
"reload <skill> config" / "reload <skill> lexicon" — re-read the named skill's config or lexicon after editing.

What stays plain English (always)

Anything inside backticks: `foo()`, `null`, file paths, flags. Hard rule, not configurable.
Anything covered by an enabled preserve.* toggle in the active skill's config.
Slash command output and harness messages (those aren't Claude's prose anyway).
Replies in non-English languages — neither the Bard nor the buccaneer spoke anything else.

Slash commands (Claude Code plugin)

When installed via the marketplace, chrysippus exposes these commands (auto-namespaced by Claude Code as /chrysippus:<command>):

Command	Effect
`/chrysippus:shakespeare`	Activate the shakespeare persona for the session
`/chrysippus:pirate`	Activate the pirate persona
`/chrysippus:gen-alpha`	Activate gen-alpha
`/chrysippus:toronto-mans`	Activate toronto-mans (hard-lock on safety warnings)
`/chrysippus:personas`	List installed personas and show which is active

Pass a flavor as an argument to switch flavor on activation: /chrysippus:pirate shanty.

Reviewer subagents

Four per-persona code-review subagents, plus a meta "dramaturg":

Agent	Purpose
`shakespeare-reviewer`	PR review in Bardic register, substantive feedback
`pirate-reviewer`	PR review in pirate register
`gen-alpha-reviewer`	PR review in gen-alpha voice
`toronto-mans-reviewer`	PR review in toronto-mans register
`dramaturg`	Meta-agent: audits persona rule-adherence in a conversation; speaks plainly itself

Each reviewer declares the corresponding skill in its frontmatter, so the persona's full rules load automatically when the subagent runs.

Invoke with an @-mention:

@chrysippus:shakespeare-reviewer review my latest PR

See docs/subagents.md for the full invocation guide — natural-language, session-default, passing context, when to use a subagent vs. your main session, and an example of what each reviewer's output looks like.

SessionStart hook (opt-in auto-activation)

hooks/session-start.sh reads .claude/personas/active in the project dir on every session start. If the file names a valid persona, the hook tells Claude to auto-activate it for the session.

Enable for a project:

bash hooks/activate.sh shakespeare   # or pirate / gen-alpha / toronto-mans
bash hooks/activate.sh status        # check current state
bash hooks/activate.sh off           # disable

Security: the hook runs unsandboxed at your shell privilege (per Claude Code's hook model). It contains no network calls, no eval, and under 100 lines; read it before running. See SECURITY.md for the full disclosure.

Multi-agent support

As of v0.5.0, the same canonical rules at rules/<persona>/ are rendered into formats for multiple AI coding agents. scripts/render.py emits all targets; scripts/render.py --check verifies no drift.

Agent	Generated file(s)	Status
Claude Code	`.claude/skills/<persona>/SKILL.md` + `<persona>.config` + `examples.md` + `lexicon.md`	✅
Codex / Cline / Aider (AGENTS.md convention)	`AGENTS.md`	✅
Gemini CLI	`GEMINI.md`	✅
Aider (project conventions)	`CONVENTIONS.md`	✅
Windsurf (Codeium)	`.windsurfrules`	✅
Cline	`.clinerules`	✅
Cursor IDE	`.cursor/rules/<persona>.mdc`	✅
Kiro IDE	(coming soon — awaiting stable spec)	⏳

All multi-agent files are checked into the repo alongside the Claude Code skills. Downstream users get activation in their preferred agent without needing to regenerate anything.

File layout

.claude/skills/
├── shakespeare/
│   ├── SKILL.md            # the skill: register, flavors, preservation rules
│   ├── shakespeare.config  # per-repo settings
│   └── examples.md         # extended before/after corpus
├── pirate/
│   ├── SKILL.md
│   ├── pirate.config
│   └── examples.md
├── gen-alpha/
│   ├── SKILL.md
│   ├── gen-alpha.config
│   ├── examples.md
│   └── lexicon.md          # current vocabulary, refreshable
└── toronto-mans/
    ├── SKILL.md
    ├── toronto-mans.config
    ├── examples.md
    └── lexicon.md          # vocabulary w/ per-term provenance (MTE, Patois, AAVE)

At the repo root, generated multi-agent files also live alongside the Claude Code skills:

AGENTS.md              # universal (Codex, Cline, Aider, others)
GEMINI.md              # Gemini CLI
CONVENTIONS.md         # Aider project conventions
.windsurfrules         # Windsurf (Codeium)
.clinerules            # Cline
.cursor/rules/<persona>.mdc  # Cursor IDE (one per persona)

All of the above and .claude/skills/<persona>/ are generated from canonical source at rules/<persona>/ by scripts/render.py. To modify a skill, edit the rules/ source and re-run scripts/render.py. See CONTRIBUTING.md for the workflow.

Credits

Voices borrowed from a man who has been dead 410 years and still writes better than most of us; from a fictional register that never really existed but has outlived most that did; and from internet-meme culture, parodied by people too old to be representative of it.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

chrysippus

Quickstart

Demo

Sibling skills

Install

Claude Code marketplace install (recommended for Claude Code users)

Per-repo install (copy the skill folders by hand)

User-global install (just you, every repo)

Activation

1. Trigger phrase (no setup)

2. Always-on via CLAUDE.md (recommended for dedicated repos)

Configuration

Shakespeare

Pirate

Gen-Alpha

Toronto-mans

Mid-session controls

What stays plain English (always)

Slash commands (Claude Code plugin)

Reviewer subagents

SessionStart hook (opt-in auto-activation)

Multi-agent support

File layout

Credits

About

Uh oh!

Releases 7

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 50 Commits
.claude-plugin		.claude-plugin
.claude/skills		.claude/skills
.cursor/rules		.cursor/rules
.github		.github
agents		agents
commands		commands
docs		docs
evals		evals
hooks		hooks
rules		rules
scripts		scripts
tests		tests
.clinerules		.clinerules
.gitattributes		.gitattributes
.gitignore		.gitignore
.windsurfrules		.windsurfrules
AGENTS.md		AGENTS.md
CLAUDE.md		CLAUDE.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
CONVENTIONS.md		CONVENTIONS.md
GEMINI.md		GEMINI.md
LICENSE		LICENSE
NOTICE		NOTICE
README.md		README.md
RELEASING.md		RELEASING.md
SECURITY.md		SECURITY.md
mkdocs.yml		mkdocs.yml

Uh oh!

Folders and files

Latest commit

History

Repository files navigation

chrysippus

Quickstart

Demo

Sibling skills

Install

Claude Code marketplace install (recommended for Claude Code users)

Per-repo install (copy the skill folders by hand)

User-global install (just you, every repo)

Activation

1. Trigger phrase (no setup)

2. Always-on via CLAUDE.md (recommended for dedicated repos)

Configuration

Shakespeare

Pirate

Gen-Alpha

Toronto-mans

Mid-session controls

What stays plain English (always)

Slash commands (Claude Code plugin)

Reviewer subagents

SessionStart hook (opt-in auto-activation)

Multi-agent support

File layout

Credits

About

Topics

Resources

License

Code of conduct

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 7

Uh oh!

Contributors

Uh oh!

Languages