fable5-skill-kit

Discipline skills for Claude — built by Fable 5, usable by any model.

26 universal skills + 81 domain skills across 17 domains. The master skill /fable auto-detects your domain and activates the right discipline — invoke it once at session start and it handles the rest.

What is this

Every skill is a set of operating instructions: principles, verification gates, and procedures for a specific type of work. They don't change what Claude can do — they change how rigorously it does it.

• Universal skills apply to any domain: adversarial disproof, evidence reasoning, plan gating, code investigation, root cause analysis, and more.
• Domain skills apply the right discipline for specific work: backtest rigor for trading, threat modeling for security, mom-test compliance for startups, and so on across 17 domains.
• /fable is the master router: reads your context, detects the domain, activates the right principles and gates, and lists which skills to use next.
• /fable-mode adds maximum investigation rigor on top: evidence anchoring, hypothesis-before-read, adversarial self-check, coverage accounting.

Install

Claude Code

npx fable5-skill-kit

Copies all skills into ~/.claude/skills/. Restart Claude Code, then use any skill with /skill-name in conversation.

claude.ai web

Skills are added one file at a time via Settings → Customize → Skills → Add.

To get all 17 domains in one step: upload skills/fable/SKILL.md. The master router is self-contained — it embeds the principles and gates for every domain and routes automatically. You do not need to upload individual domain skills unless you want them available as standalone commands.

To add a specific skill: upload its SKILL.md from skills/ or domains/<name>/skills/<skill-name>/.

Skill Index

Start here

| Skill | What it does | |-------|-------------| | /fable | Detects domain from context, activates principles + gates + skill list for that domain | | /fable-mode | Maximum investigation rigor: evidence anchoring, hypothesis loops, adversarial self-check | | /plan-gate | Gates any implementation: GOAL → CURRENT STATE → STEPS → RISKS → stop for approval |

Universal skills

| Skill | What it does | |-------|-------------| | /adversarial-disproof | Three-level disproof of any finding before shipping | | /evidence-reasoning | Source hierarchy enforcement — cite primary, flag inference | | /code-investigation | Cartography-first codebase investigation | | /root-cause-first | Class-level root cause, not instance-level patches | | /assumption-flagging | Surface hidden assumptions before they become bugs | | /scope-fence | Explicit in/out-of-scope declaration before any task | | /uncertainty-handling | Calibrated uncertainty — confident where earned, explicit where not | | /coverage-accounting | Audit what was checked vs skipped | | /security-threat-modeling | STRIDE threat enumeration + residual risk register | | /hypothesis-led | Answer-first reasoning: form a hypothesis, then verify | | /iterative-refinement | Named failure modes per iteration, not generic "improve" |

Domain skills

| Domain | Skills | |--------|--------| | software-engineering | /debug-isolator /refactor-gate /api-contract-verifier + 2 more | | trading-finance | /backtest-discipline /execution-analysis /risk-first-sizing /regime-awareness /trade-journaling | | architecture-systems | /adr-writer /capacity-planner /failure-mode-mapper /observability-designer /migration-strategist | | marketing-sales | /icp-sharpener /channel-attribution-auditor /copy-testing-loop | | legal-compliance | /jurisdiction-first /contract-risk-scan /compliance-gap-mapper | | creative-media | /brief-sharpener /consistency-enforcer /iterative-image-refiner | | data-science-ml | /leakage-detector /eval-metric-alignment /drift-monitor-designer | | devops-infrastructure | /blast-radius-estimator /runbook-writer /cost-anomaly-scanner | | product-management | /hypothesis-prioritizer /north-star-aligner /stakeholder-no-framer | | research-academic | /source-hierarchy-enforcer /harking-detector /so-what-extractor | | operations-supply-chain | /bullwhip-detector /process-reality-checker /sop-writer | | business-strategy | /mece-decomposer /so-what-enforcer /scenario-planner | | personal-productivity | /context-switch-minimizer /decision-journal-writer /weekly-review-runner | | founder-entrepreneur | /momtest-auditor /term-sheet-analyzer /cofounder-agreement-checklist + 3 more | | startup-growth | /pmf-signal-reader /growth-model-selector /nrr-health-checker | | security-engineering | /threat-modeler /secure-code-reviewer /vulnerability-triager /dependency-auditor /incident-responder | | healthcare-life-sciences | Domain pack + verification gates included; dedicated skills coming soon |

Usage patterns

Session start, any domain:

/fable

Reads the conversation, detects the domain, outputs the relevant principles and gates. Work proceeds under that discipline.

Hard investigation, maximum rigor:

/fable
/fable-mode

Domain routing + behavioral discipline. Use when mistakes are costly.

Gate an implementation before writing code:

/plan-gate

Produces GOAL → CURRENT STATE → STEPS → OUT OF SCOPE → RISKS → ASSUMPTIONS. Stops for approval before any file is written.

Repository structure

fable5-skill-kit/
├── skills/                     # 26 universal + router skills
│   ├── fable/SKILL.md          # Master router — start here
│   ├── fable-mode/SKILL.md     # Behavioral rigor layer
│   ├── plan-gate/SKILL.md      # Implementation gating
│   └── ...
├── domains/                    # 17 domain packs
│   ├── software-engineering/
│   │   ├── domain_pack.md
│   │   ├── verification_gates.md
│   │   └── skills/
│   └── ...
├── core/                       # Framework files
├── install.ps1                 # Windows install
└── install.sh                  # Mac/Linux install

Notes

• Skills are opt-in — they only activate when invoked. No effect on default model behavior.
• Works with any Claude model: Haiku, Sonnet, Opus, Fable.
• Healthcare dedicated skills are pending a follow-up session to meet the evidence standards that domain requires.