immortal-skill
by agenmodimmortal-skill is a Skill Authoring workflow for turning chats, documents, and platform exports into an evidence-aware digital persona package. It guides role selection, source collection, ethics checks, conflict handling, and output generation for reusable agent skills.
This skill scores 72/100, which means it is listable for directory users as a real, reusable workflow, but adoption will still require some repository reading and manual setup. The repository shows a substantive distillation framework with clear activation cues, phased process structure, role templates, output path conventions, and platform-specific collection guidance, so an agent has more to work with than a generic prompt. The main limitation is that the primary skill file appears instruction-heavy and references many supporting paths without giving one compact end-to-end quick start inside the skill itself.
- Clear triggerability: SKILL.md explicitly states when to activate for requests like distilling a person into a digital persona/skill package.
- Substantive workflow depth: the main skill defines phased steps, ethics checks, role templates, evidence handling, and output conventions instead of staying at a marketing level.
- Repository has real operational support: file tree includes collectors for many platforms and docs/PLATFORM-GUIDE.md with concrete setup and collect commands for tools like Feishu, Slack, Discord, and Telegram.
- Primary install/use path is fragmented: SKILL.md references personas, recipes, kit, and docs, but the evidence provided does not show a single concise end-to-end command sequence in the main skill.
- Trust and fit are sensitive: the repository addresses ethics, privacy, and public/private source limits, but the subject matter involves distilling real people, so users need to apply those safeguards carefully.
Overview of immortal-skill skill
What immortal-skill is actually for
immortal-skill is a structured skill for turning scattered records about a person into an agent-loadable “digital persona” package. Its real job is not simple chat-log summarization. It guides an agent through role selection, source collection, evidence grading, conflict handling, and output generation so the result captures how someone thinks, communicates, remembers, and decides.
Who should install immortal-skill
The best fit for immortal-skill is anyone doing serious Skill Authoring around a person-shaped knowledge asset: preserving your own working style, documenting a mentor, retaining team know-how from a departing colleague, or building a public-figure advisor from public materials. It is especially useful when ordinary prompting feels too loose and you need a repeatable structure across multiple cases.
Why users pick this over a generic prompt
The main differentiator is workflow depth. The repository includes persona templates, platform-specific collectors, examples, recipes, and two related protective skills: distill-protocol-skill for usage terms and distill-shield-skill for defensive canaries. That makes immortal-skill skill more operational than “paste messages and ask the model to imitate someone.”
Biggest adoption caveats first
immortal-skill is conceptually strong, but adoption depends on your data access and ethics constraints. Many useful sources require platform tokens, exports, or local databases. The repo expects python3, and practical collection quality varies by source. If you only want a quick style mimic from a few paragraphs, this is probably too heavy; if you need traceable, reusable distillation, it is a better fit.
How to Use immortal-skill skill
immortal-skill install and first files to read
Start by opening SKILL.md first, then FOR_AI.md, README.md, docs/PLATFORM-GUIDE.md, and the relevant files under personas/ and recipes/. The repository does not present a single copy-paste install command in SKILL.md, so treat this as a repo-backed workflow skill rather than a package with one launcher. You will likely need python3 if you plan to use collectors or helper scripts.
Read in this order:
SKILL.mdfor the canonical workflowpersonas/_base.mdplus one role templaterecipes/intake-protocol.mddocs/PLATFORM-GUIDE.md- related subskills if permissions or protection matter
Inputs immortal-skill needs to work well
For good immortal-skill usage, give the agent four things up front:
- the target person and role type: self, colleague, mentor, family, partner, friend, or public figure
- allowed data sources and collection limits
- the output goal: archive, internal training, advisor, memory preservation, or agent skill pack
- ethics boundaries: consent, redaction, public-only limits, family sensitivity, or workplace scope
A weak request is: “Distill my mentor.”
A strong request is: “Use immortal-skill to distill my mentor into a Skill Authoring package for internal coaching. Sources: exported email threads, 12 meeting notes, and public talks only. Redact third-party private details. Prioritize teaching heuristics, feedback style, and decision principles over biography.”
Practical workflow from rough goal to usable output
A reliable immortal-skill guide looks like this:
- Choose the role template before collecting data.
- Confirm ethics rules for that role.
- Gather material from one or two high-signal sources first, not every platform at once.
- Distill by dimensions the skill emphasizes: procedure, interaction, memory, personality.
- Ask for evidence levels and conflict notes instead of a single merged voice too early.
- Generate the final skill package under
./skills/immortals/<slug>/.
This matters because early over-merging creates polished but untrustworthy personas. Better results come from separating “what is clearly supported” from “probable pattern inferred from multiple sources.”
Collection paths and quality tips that change results
The repo supports many inputs through collectors/ such as slack.py, discord.py, telegram.py, wechat.py, email_collector.py, and manual.py. If you can automate collection, check docs/PLATFORM-GUIDE.md first for token scopes and setup steps. If not, use manual curation and keep source provenance.
Quality tips:
- prefer longer, contextual exchanges over isolated quotes
- include examples of decisions, feedback, conflict, and routine behavior
- separate public materials from private materials
- label time periods if the person changed noticeably over time
- for public figures, stay strictly source-traceable
immortal-skill skill FAQ
Is immortal-skill good for beginners?
Yes, if you can follow a staged workflow. immortal-skill is easier than inventing your own distillation method, but it is not “one prompt and done.” Beginners should start with self or public-figure cases because consent and source boundaries are clearer.
When is immortal-skill better than ordinary prompting?
Use immortal-skill skill when you need consistency, evidence-aware outputs, and a reusable artifact. A normal prompt may be enough for “sound a bit like this person.” immortal-skill is better for preserving methods, communication patterns, and decision rules in a way another agent can load later.
When should you not use immortal-skill?
Skip it if you lack permission, only have tiny amounts of low-context data, or just want a lightweight character simulation. It is also a poor fit when the real need is factual knowledge extraction rather than person-level distillation.
Does immortal-skill fit a broader workflow?
Yes. The repo is designed as a small ecosystem. Use distill-protocol-skill if you need usage terms or machine-readable restrictions, and distill-shield-skill if you want canaries or protective markers before handing over data. That makes immortal-skill for Skill Authoring stronger in real organizational use.
How to Improve immortal-skill skill
Give immortal-skill better source material
The fastest improvement is not prompt cleverness; it is source selection. For immortal-skill, the best corpus usually mixes:
- natural conversations
- artifacts showing judgment, like emails or reviews
- documents showing process
- a few emotionally or strategically important moments
Avoid giant unfiltered dumps with no metadata. Ten labeled, representative samples often beat thousands of raw lines.
Ask for evidence-aware drafts, not final personas first
A common failure mode is asking for a finished “digital twin” too early. Instead, tell the agent to produce:
- extracted traits by dimension
- supporting evidence snippets
- contradictions or uncertainty
- only then a merged persona draft
This improves trust and makes later edits much easier.
Tighten prompts around role, audience, and boundaries
Better immortal-skill usage comes from explicit constraints. Say who will use the output, what it should help them do, and what it must avoid. Example: “Write for internal onboarding, not companionship. Preserve blunt feedback style, but remove private family references and identifiable third-party details.”
Iterate after the first draft like an editor
After the first pass, do targeted revision rounds:
- “What claims lack evidence?”
- “Which traits are overfit to one source?”
- “What is missing about decision-making under pressure?”
- “Rewrite the skill package to sound less flattering and more operational.”
That is where immortal-skill becomes valuable: not as a novelty clone, but as a disciplined distillation workflow you can inspect, improve, and trust.