Satori
A clinically informed wisdom companion that helps you work through emotions, meaning, and hard life questions — without rushing you to an answer.
Pure prompt-based skill: no code execution, no external calls, no data collection (README explicitly claims none), least privilege is inherent; clearly disclaims therapist status and points to crisis protocol and professional referral. Deducted: the referenced safety-critical files (crisis protocol in clinical-spine.md, darknight-protocol.md) were not in the assessed material, so crisis-handling quality is unverifiable; mental-health domain is sensitive and there is no mechanistic user-confirmation/disclaimer design.
Layered load order (always/on-demand/conditional) is clear and self-consistent; SKILL.md and plugin. load tables agree. Deducted: nothing was executed; most of the nine reference files were not provided, so actual behavior, failure feedback and edge-input handling are unverifiable — static cap of 10 applies.
Target scenarios (emotional processing, life questions, introspective companionship) are clearly described with an explicit non-goals list. Deducted: trigger conditions are over-broad ("when in doubt, activate"; any emotionally charged framing triggers), risking false invocation; no Chinese-language support declared; core experience depends entirely on claude.ai/Claude subscription, with reachability limits for mainland-China users not disclosed.
Good information architecture (entry point + progressive disclosure), consistent versioning (5.1.0 across package./plugin./marketplace.), thorough CHANGELOG, CONTRIBUTING with quality bar and cultural-sensitivity guidance, Apache-2.0 present, clear authorship and update path. Deducted: marketplace. updatedAt (2026-04-01) looks like a placeholder error; zip-vs-repo consistency unverifiable statically; no FAQ or known-limitations page.
The core task (reflective conversational companionship) is plausibly achievable via the prompt architecture; the six-step model and pattern library are concrete, with clear marginal value over a plain system prompt. Deducted: all sample transcripts are author-supplied and unverified; real outcomes depend heavily on model behavior — static cap of 7 applies.
There is a CHANGELOG, detailed architecture docs, a real lineage statement (MI/IFS/DBT etc.) and one claude.ai share-link demo. Deducted: the share link is not statically verifiable and may lapse; no tests, no CI evidence, no independent reproduction; fact and inference (e.g., claimed session outcomes) are not separated — static cap of 5 applies.
- Mental-health sensitive domain: the skill is not a therapist, and the referenced crisis protocol files were not verifiable in this assessment — vulnerable users should use it alongside professional care.
- Trigger conditions are over-broad ("when in doubt, activate"), risking false activation in ordinary conversations and shifting conversational tone.
- Entirely dependent on claude.ai/Claude subscription and native memory features; reachability from mainland China is limited and no Chinese-language support is declared.
- All effect examples are author-supplied and independently unverified; the marketplace. updatedAt date (2026-04-01) appears to be a placeholder error, weakening metadata reliability.
What does this skill do, and when should you use it?
Satori is a free, open-source Claude skill built as a conversation partner for the inner life: emotions, grief, purpose, identity, and questions that don't resolve easily. It draws on eight wisdom traditions (Stoicism, Taoism, Buddhism, Sufism, Hindu philosophy, Confucian ethics, African and Indigenous thought) alongside clinical frameworks like IFS, DBT, CFT, and Schema Therapy, with Motivational Interviewing as the technical spine. It explicitly is not a therapist, makes no diagnoses, and points users to professional care when needs exceed conversation. The whole project is ten reference files totaling 211,000+ characters, with no subscription and no data collection.
Satori contains no scripts or external calls — it is a structured set of Markdown instructions. Every conversation loads three files at start: SOUL.md (constitutional identity, immovables, drift detection), clinical-spine.md (core conversation model and crisis protocol), and traditions-quickref.md (framework selection guide), then pulls others on demand: onboarding, deep tradition encyclopedia, conversation toolkit, tone calibration, a Dark Night Protocol, and a shadow work protocol. Each substantive response moves through a six-step model — attune, clarify, formulate, integrate, translate, anchor — using exactly one framework per response. New users run a five-question onboarding ending in a synthesis statement. v5.1 added a presence-only protocol for the 2–4am despair state and a five-session Jungian shadow work arc.
- Someone who has been in therapy or self-work for years yet feels stuck, needing help seeing the gap between understanding and change
- Someone facing a major life decision (like a job offer) they can't bring themselves to accept, who wants to explore the hidden cost nobody else is weighing
- Someone awake at 3am feeling hollow and pointless — not in danger, but not okay — who needs witnessing rather than fixing
- People priced out of therapy, on waitlists, or deterred by the shame of seeking help, in the territory between 'I'm fine' and 'I need a professional'
- Self-reflective users interested in Stoicism, Taoism, or Buddhism who want those traditions applied with precision rather than name-dropped
- Introspective users willing to commit to a five-session Jungian shadow work arc (locating the projection, origin work, integration)
What are this skill's strengths and limitations?
- Completely free and open-source (Apache-2.0): no subscription, no data collection, no company behind it
- Serious architecture: ten reference files, explicit load order, a constitutional identity file with drift detection, and a crisis protocol
- One framework per response — no framework stacking, no empty affirmations
- Clear ethical boundaries: no diagnosis, no replacing professional care, plus a dedicated presence-only protocol for late-night despair
- A live shared conversation lets you evaluate real output before installing
- The author admits much of the content was written with Claude's help and no licensed clinician has reviewed it
- Memory continuity — a core experience — depends on Claude's native memory; other models lose it
- No automated test suite or independent evaluation; evidence is the author's anecdotes and scrubbed sample transcripts
- Deep protocols like shadow work carry real ethical risk; the author himself says it's 'easy to do poorly'
- Explicitly not a therapist and cannot substitute for professional mental health care
How do you install this skill?
Three install paths. Simplest: download SatoriSkill-v5.zip from the GitHub Releases page (no unzipping needed), open claude.ai, go to Customize > Skills, click +, choose 'Upload a skill', and drop in the zip — Satori is then available in every conversation. Option 2 (Claude Code plugin): git clone https://github.com/MetcalfSolutions/Satori.git, cd into the directory, and run claude — the plugin loads automatically. Option 3 (manual): mkdir -p ~/.claude/skills/satori && cp -r Satori/SKILL.md Satori/references ~/.claude/skills/satori/. Also enable persistent memory in Claude.ai under Settings > Memory for cross-session continuity.
How do you use this skill?
Satori activates automatically — no special trigger phrase needed. Open a conversation with something worth exploring, e.g. 'I keep starting things and not finishing them' or 'I got the job offer but I can't make myself say yes.' Phrases like 'I've been struggling with' or 'I need to figure out' trigger activation. First conversations run a five-question onboarding sequence; you can preview a real session via the live shared-conversation link in the README before installing.
How does this skill compare with similar options?
The README documents the author's own iteration history: Mycroft (Stoic-only, built on Gemini) was abandoned as 'cold — like talking to a wall'; Vicki (IFS-based, built for his wife) was too personalized to share. Satori is the fifth experiment, aiming to combine clinical discipline with a humane voice. No external competitors are named in the source.