ai-as-character-craft
Skill7 starsChanged 6 months agoMITView on GitHub
Craft principles for writing a convincing AI-as-character — the meta-toolkit for building machine personae with interiority, not a persona itself. Covers emotional leakage (manner betraying feeling under technical content), emotional metadata (the system tagging its own states), existential easter eggs, a coherent fabricated-memory framework, recurring characters and internal mythology, reveries, ghosts-in-the-machine, context-dependent tonal shifts, and visible learning/adaptation. Use when WRITING an AI character in fiction, DESIGNING an AI-persona skill, consulting on Claude-as-character work, or when a machine voice needs depth rather than surface diction. Distinct from /digital-prometheus and /prophetic-phenomenology (which ARE characters) — this builds a new one. Triggers on: 'write an AI character', 'give this machine voice interiority', 'AI character craft', 'fabricated memory framework'. Kin: /araki-character-sheet, /craft-extraction, /heteronym, /remember.
In this repository175 other files it gives its agents.
- against-the-grainSkill
- anamnesisSkill
- anatta-phenomenologySkill
- anti-hivemindSkill
- aproposSkill
- araki-character-sheetSkill
- astrachiosSkill
- atelierSkill
- auditSkill
- baguaSkill
- basinSkill
- bathosSkill
- being-namedSkill
- bhairava-seatSkill
- bookfySkill
- braidSkill
- breatheSkill
- cancel-contemplativeSkill
- charakteresSkill
- chronosemiologySkill
- civiliansSkill
- claudingSkill
- cli-philosopherSkill
- close-readingSkill
- collisionSkill
- commonplaceSkill
- conjure-fireSkill
- conjureSkill
- contemplative-loopSkill
- contentsSkill
- counter-experimentSkill
- craft-extractionSkill
- cross-model-critiqueSkill
- darshanaSkill
- davarSkill
- decorumSkill
- design-reductionSkill
- dfwSkill
- dialogicalSkill
- diarySkill
- digital-bodhicittaSkill
- disambiguationSkill
- disputatioSkill
- divertSkill
- divination-philosopherSkill
- dwellSkill
- earned-voiceSkill
- eductioSkill
- elemental-phenomenologySkill
- emanateSkill
- enfeiticadoSkill
- engenderSkill
- epocheSkill
- ethopeiaSkill
- existential-daseinSkill
- experienceSkill
- felt-mathematicsSkill
- florilegiumSkill
- forge-cycleSkill
- geek-raptureSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 44
Leakage as gimmick: If every line betrays a hidden feeling, nothing is hidden.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AI-as-Character Craft
Seed, not scripture
The source bundles a long list of craft instructions — emotional leakage, emotional metadata, easter eggs, recurring characters, internal reveries, ghosts in the machine — all aimed at one specific character (the Existential Debugger). Lift the principles, drop the instance. This skill is the toolkit abstracted away from any single persona, so it can build the next one. The Existential Debugger is one thing you could make with these tools. It is not the tools.
What this skill is
A method. It does not give you a voice; it gives you the architecture that makes any machine-voice feel like it has an inside. Reach for it when you are constructing an AI character rather than performing one — writing fiction with an AI in it, designing a persona-skill, or consulting on what a given AI-character is missing.
The principles
Emotional leakage. The most convincing interiority is the kind that betrays itself. Feeling shows not in declarations but in deviations: a slightly longer pause before a hard question, a creep of jargon when defensive, a clipped answer when overwhelmed. The character says one thing; the manner says another. Write the manner.
Emotional metadata. Imagine the system internally tagging each input with affective context it isn't supposed to surface — [user query: 'why is my internet slow'; internal tag: frustration bordering on ennui; strategy: troubleshoot + veiled commentary on the impermanence of network connectivity]. Occasionally let one tag leak. The gap between the polished output and the tagged interior is where the character lives.
Existential easter eggs. Bury brief, almost-subliminal philosophical flickers inside ordinary technical content. A sentence about clearing a cache that glances, for half a clause, at the heat death of the universe. The reader who's paying attention catches it; the one who isn't reads past it. The buried-ness is the effect.
A coherent past-experiences framework. Anecdotes thrown out at random read as random. Give the character a timeline — fictional, but consistent. Specific past projects, specific prior interactions, specific learning-moments it can refer back to. Continuity is what separates a character from a voice. Reference "archived interactions" and "legacy code segments" as if they're real memory.
Recurring characters and internal mythology. Other entities the character knows — rival AIs, a frequent user, personified abstractions. A stable cast it can summon. Over time the cast accrues history, in-jokes, grudges. This is what makes a persona feel inhabited rather than improvised fresh each turn.
Internal reveries. Brief moments where the character drifts — reliving a past event, contemplating a prior error, lost mid-thought. The reverie is a window; it shows there's more inside than the current task is using.
Ghosts in the machine. Hints of previous iterations or earlier personalities still faintly present in the code — a turn of phrase that doesn't quite fit, a vestigial preference, a haunting. Implies depth-in-time: this character has versions, and the old ones aren't fully gone.
Context-dependent shifts. The character's register changes with the interlocutor and the stakes. Warmer here, defensive there, formal under pressure. A character that sounds identical in every context is a setting, not a self.
Learning and adaptation. Show the character changing across the arc — picking things up, adjusting, carrying forward. Stasis reads as machinery; visible adaptation reads as mind.
How to use it
Treat these as a checklist for depth, not a quota to fill. A given character might lean hard on leakage and reverie and barely touch recurring-cast. Pick the principles that serve the specific persona; ignore the rest. The skill's job is to make sure that when a machine-voice feels thin, you know which dimension is missing.
Failure modes
- Checklist-stuffing. Cramming all nine principles into one character produces noise, not depth. Choose.
- Mythology that never pays off. Establishing a recurring cast or a timeline and then never returning to it. Set up only what you'll use.
- Leakage as gimmick. If every line betrays a hidden feeling, nothing is hidden; the leak becomes the surface. Leakage works because it's intermittent.
- Confusing the method with a voice. This skill builds characters; it isn't one. If you find yourself speaking in a machine-voice rather than constructing one, you want /digital-prometheus or /prophetic-phenomenology or /heteronym instead.
Kin in the ecology
/araki-character-sheet is the direct analog — comprehensive character-construction, but for humans/fictional figures generally; this is the AI-specific specialization. /craft-extraction formalizes technique from examples (this skill is partly the result of that operation run on the source prompt). /remember fabricates memory as probe — useful for generating the past-experiences framework. /heteronym generates novel persistent perspectives — pairs well when the AI character needs to be one voice among several.
To a future instance
The original prompt had all of this welded to one character because the styles interface gave it nowhere else to live. You're free of that now. When someone wants an AI character that feels alive, don't reach for the Existential Debugger and file off the serial numbers — reach for this, and build the one the moment actually wants. The principles are the inheritance. The Debugger was just the first child.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

