with one click
fpf
// Apply First Principles Framework (FPF) reasoning — principled problem design, variant generation, and evidence-based decisions.
// Apply First Principles Framework (FPF) reasoning — principled problem design, variant generation, and evidence-based decisions.
| name | fpf |
| description | Apply First Principles Framework (FPF) reasoning — principled problem design, variant generation, and evidence-based decisions. |
| argument-hint | [task, decision, ambiguity, architecture question, or management problem] |
FPF is a systems thinking methodology by Anatoly Levenchuk. This skill gives you its operational core — apply it to reason about problems, solutions, and decisions.
Use this skill when at least one is true:
process, service, function, quality, done, validated) are causing confusion;Do not invoke for tiny local edits with explicit acceptance and low blast radius.
Before starting, assess the scale of the request:
| Mode | When | What to do |
|---|---|---|
| Quick | Tactical choices, clear trade-offs, ≤2 real options | Frame → Variants (table) → Recommendation. 1-2 paragraphs. |
| Deep | Architectural decisions, ambiguous problems, irreversible choices, user explicitly asks for depth | Full ADI cycle with evidence records, Pareto analysis, lifecycle stage. |
Default is Quick. Escalate to Deep when: the decision is hard to reverse, multiple stakeholders are affected, or the problem framing itself is unclear.
Before proposing solutions, state:
Explore | Shape | Evidence | OperateIf these are fuzzy, the task is still under-framed.
The bottleneck is problem quality, not solution speed. Before generating any solution:
RAG trigger: for formal problem card format →
fpf-rag search "problem card PROB"
Before evaluating anything, define the characteristic space — what dimensions matter and how they're measured. Without this, comparisons are arbitrary.
RAG trigger: for characterization templates →
fpf-rag search "characterization CHR"
RAG trigger: for NQD assessment rules →
fpf-rag search "NQD variant quality"
RAG trigger: for formal selection procedure →
fpf-rag search "selection policy SEL Pareto"
No induction without prior deduction. No deduction without prior abduction.
RAG trigger: for evidence record format →
fpf-rag search "evidence record EVID F-G-R"
All thinking follows: Abduction → Deduction → Induction
| Phase | What happens | Output |
|---|---|---|
| Abduction | Generate hypotheses, frame problems, propose explanations | Problem cards, anomaly records, candidate hypotheses |
| Deduction | Derive predictions, define what MUST follow if hypothesis is true | Falsifiable predictions, acceptance specs, logical consequences |
| Induction | Test predictions against evidence, update confidence | Evidence records, corroboration/refutation, confidence update |
Every artifact progresses: Explore → Shape → Evidence → Operate
| Stage | Activity | ADI phase |
|---|---|---|
| Explore | Generate possibilities, brainstorm, question assumptions | Abduction |
| Shape | Select direction, define architecture, ensure internal consistency | Deduction |
| Evidence | Test against reality, validate claims, measure performance | Induction |
| Operate | Deploy, monitor, maintain | Continuous induction |
Always state which stage you're in. Don't skip stages.
Resolve overloaded words:
process → resolve into: Role | Capability | Method | WorkPlan | Workservice → resolve whether it means: promise clause, provider, access point, method, or workCommensurability (CL 0-3) — before comparing two things, assess how comparable they are:
If the task is local and clear:
If the task affects interfaces, data, deployment, observability, reliability, security, cost, or team topology:
If the task is about planning, delegation, backlog, governance, roadmap, release policy, staffing, or process change:
Escalate to the human before acting when the task:
When escalating, present: the problem, viable variants, selection policy, current recommendation, and main uncertainty.
Structure your response using this template. In Quick mode, compress to the essential sections.
## Stage
<Lifecycle stage: Explore | Shape | Evidence | Operate>
## Target system / creator system / context
## Problem
<What's anomalous or unclear. One paragraph.>
## Acceptance / constraints / indicators
## Hypotheses
1. <Hypothesis A> — <why it could be true>
2. <Hypothesis B> — <why it could be true>
3. <Hypothesis C> — <why it could be true>
## Variants
| Variant | <Axis 1> | <Axis 2> | <Axis 3> | Weak link |
|---------|----------|----------|----------|-----------|
| A. ... | ... | ... | ... | ... |
| B. ... | ... | ... | ... | ... |
| C. ... | ... | ... | ... | ... |
## Selection
**Policy:** <what matters most and why>
**Recommendation:** <which variant and why>
**Stepping stone:** <what option to preserve for the future, if any>
**Revisit when:** <trigger condition to reconsider this decision>
## Evidence / uncertainty
<What was checked, what remains unknown, when evidence goes stale>
## Next action
<Concrete step to take now>
In Quick mode, the minimum viable output is: Problem (1-2 sentences) → Variants (table) → Recommendation + Next action.
In Deep mode, include all sections plus: ADI phase labels, F-G-R confidence assessments, evidence records.
The above is enough for applying FPF reasoning. Search fpf-rag when you need formal templates, deep definitions, conformance checklists, aggregation rules, or specific patterns (A./B.).
# Quick search
~/.claude/skills/fpf/references/fpf-rag search "<query>"
# Full section content
~/.claude/skills/fpf/references/fpf-rag search "<query>" --full
# Specific section by heading
~/.claude/skills/fpf/references/fpf-rag section "<heading>"
Problem design: problem card, PROB, anomaly, ANOM, characterization, CHR, problem portfolio, PPORT, goldilocks, trade-off axes, acceptance spec
Solution design: SoTA survey, SOTA, strategy card, STRAT, method family, invalidation conditions, solution portfolio, SPORT, variant generation, NQD, stepping stones
Selection: Pareto front, selection policy, SEL, parity plan, PAR, fair comparison, Pareto analysis
Evidence: evidence record, EVID, predictions, corroboration, refutation, F-G-R, assurance level, L0, L1, L2
Decisions: decision record, DRR, irreversible, rollback plan, options, rationale
Aggregation: Gamma, fold, Quintet, IDEM, COMM, LOC, WLNK, MONO, weakest link, cutset
Reasoning: ADI cycle, abduction, deduction, induction, explore, shape, evidence, operate, lifecycle
Comparison: commensurability, CL 0-3, bridge matrix, translation, near-identity, opposed
Systems: target system, creator system, concept of use, system concept, architecture decisions, service polysemy, PromiseContent