learn skills
12 skills. Source of truth: skills/learn/.
feynman-loop
The core Feynman learning cycle for prometheus-skill-pack. Maps Feynman's explain-grade-gap-relearn cycle to the PMPO lifecycle. Supports vertical recursion (child loops on gap concepts), horizontal escalation (novice→peer→skeptic audiences), recursion floor guards, and all three mastery closure criteria.
v1.1.0 · learn feynman pmpo loop recursion escalation mastery · source
learn-about-system
Zero-friction adoption entry point for the Prometheus Skill Pack. Routes operators into the Feynman learning loop for the KBD lifecycle, skill pack capabilities, or AI harness orientation. The skill pack teaches itself using its own learning infrastructure.
v1.0.0 · learn adoption meta-learning kbd skill-pack harness onboarding · source
learn-certify
Certification endpoint for the Feynman learning loop. Runs prerequisite gates (feynman-artifacts, practice results, capstone), emits an Open Badges 3.0 / W3C Verifiable Credential self-issued JSON-LD signed with did-plc, and detects anomalous mastery trajectories via an integrity guardrail.
v1.1.0 · learn certify credential badge ob3 w3c-vc mastery did-plc · source
learn-goal
Entry point for the Feynman learning flow. Accepts a learning desire, assembles a grounded corpus (public and/or custom KB), runs an honest feasibility gate with sycophancy-correction, and produces a goal artifact that downstream learn-* skills consume.
v1.0.0 · learn goal feynman feasibility corpus entry-point · source
learn-grade
External, source-grounded, anti-sycophantic grader for the Feynman learning loop. Checks explanations against the teaching corpus, identifies gaps, generates novel transfer problems, and updates the learner model. Routes through sycophancy-correction to prevent pedagogical flattery.
v1.1.0 · learn grade assessment feynman sycophancy corpus-grounded transfer · source
learn-harness
Per-harness capability orientation for the Prometheus Skill Pack. Auto-detects the running AI harness via detect-surface-tier.sh, emits a capability map, and optionally routes into the Feynman loop for deeper harness understanding. Covers Claude Code, OpenCode, Codex, Kimi Code, and Zed.
v1.0.0 · learn harness claude-code opencode codex kimi zed capability-map orientation · source
learn-kb
Operator knowledge base management for the Feynman learning loop. Lets operators add, list, query, update, and remove custom knowledge bases (Dify, surreal-memory palace, local files, Firecrawl URLs) that ground the learning loop in domain-specific material. KB content never leaves the local environment.
v1.0.0 · learn kb knowledge-base dify palace grounding operator custom-knowledge · source
learn-plan
Adaptive curriculum planner for the Feynman learning loop. Reads the survey result and learner model to build a concept DAG in surreal-memory, sequences prerequisites, estimates time per concept, and produces a curriculum.json that feynman-loop and learn-practice consume. Supports --replan when mastery diverges.
v1.0.0 · learn plan curriculum dag prerequisites adaptive surreal-memory · source
learn-practice
Deliberate practice skill for the Feynman learning loop. Generates difficulty-gated, interleaved problem sets across derivation, implementation, and transfer modes. Grades via learn-grade, updates the learner model, and prevents illusion-of-knowing by requiring demonstrated performance rather than felt fluency.
v1.1.0 · learn practice deliberate-practice interleaving transfer mastery · source
learn-retain
Spaced repetition review skill for the Feynman learning loop. Reads the FSRS due queue from the learner-model crate, surfaces review prompts via ui-surface, grades retention via learn-grade at a 0.6 threshold, and updates FSRSCard state. Prevents knowledge decay after feynman-loop and learn-practice.
v1.1.0 · learn retain spaced-repetition fsrs review memory · source
learn-survey
Diagnostic placement skill for the Feynman learning loop. Generates dynamic survey items from the teaching corpus, probes current knowledge state and misconceptions, sets the recursion floor, and seeds the learner model via the learner-model crate.
v1.0.0 · learn survey diagnostic placement learner-model misconceptions · source
ui-surface
Cross-harness UI surface resolver for learn-* skills. Detects the active surface tier via detect-surface-tier.sh and renders UI intent (prompts, questions, feedback) at the best available tier. Degrades gracefully to Tier 0 text when richer surfaces are unavailable.
v1.0.0 · ui surface cross-harness learn tier-detection · source