learning_progression_ladders v1.0.0

name: "learning_progression_ladders" description: "Breaks a complex topic into a five-level progression from first contact to independent work: what changes at each level, proof tasks that show the level is real, gate checks, foundation checks that stop learners skipping basics, and remediation paths. Triggers when someone asks to sequence learning for a topic, build a curriculum ladder, define beginner to advanced levels, place a learner, or stop learners advancing on shaky foundations." metadata: { "includeInPrompt": true }

Learning Progression Ladders

Purpose

Turn one bounded topic into a five level ladder a learner can actually climb and a teacher can actually check. Each level states what the learner can do, what they cannot do yet, what support is present and how it fades, a proof task on unseen material, a pass standard, a gate check, and a remediation path for failure. Two complete ladders ship with the skill, one technical and one creative, in references/worked-ladders.md with machine readable copies in examples/.

The levels are a design choice for a bounded topic, informed by three verified models. The Dreyfus model (1980) supplies the shift from rigid rule following at novice level to holistic, prioritising work at proficient level. The SOLO taxonomy (Biggs and Collis, 1982) supplies a way to read response depth, from one aspect, to several disconnected aspects, to an integrated whole carried into a new context. The revised Bloom taxonomy (Anderson and Krathwohl, 2001) supplies verbs for pitching proof tasks, from remember and understand up to evaluate and create. Level 5 in a ladder means independent work inside the stated scope. It does not mean recognised expertise in a whole field, and the ladder must never claim it does.

Workflow

1. Bound the topic. Write the terminal performance first: what the learner does, on what material, under what conditions, without help at level 5. Write the scope boundary next, naming what the ladder deliberately leaves out. List prerequisites as things the learner can already do, not courses they sat through. If the topic cannot be bounded in two sentences, split it into two ladders.
2. Design backward. Follow the backward design order from Wiggins and McTighe: desired results first, acceptable evidence second, learning experiences last. Write all five proof tasks before writing any practice material. A level whose proof cannot be written is a level that is not defined yet.
3. Write the five levels. Use the level shapes in references/level-descriptors.md. Level 1, first contact: reproduce the core move on new material with full support. Level 2, guided use: routine work without templates, aspects handled separately and correctly. Level 3, independent routine work: plan and complete whole routine tasks, repair a planted fault by naming its cause. Level 4, adaptive work: handle the case that departs from the normal pattern, defend choices, state limits. Level 5, independent work: full task on novel material, own checks set by the learner, plus reviewing another person's work in usable terms.
4. Write one proof task per level. Each proof uses unseen or new material, never the worked example from practice. Each proof has a pass standard written as observable results, not grades or vibes: what the output contains, what the learner can explain, what check reconciles. Support is removed for proof tasks from level 2 up, in line with scaffolding practice, where temporary support inside the zone of proximal development is faded as competence grows.
5. Add the gate and the foundation check. The gate check tests the one move the next level depends on, for example explaining why a step works or finding a planted fault. Levels 2 to 5 also carry a foundation check: a small cold sample of earlier level work, mixed in without warning. A failed sample sends the learner back to re-prove that earlier level only. The whole ladder is never repeated for one gap.
6. Write practice to match the proof. Practice at each level targets one aspect at a time, with a specific goal and fast, specific feedback, following the deliberate practice pattern. Vary the surface of practice items so the learner learns the move, not the worksheet. Practice that does not resemble the proof task in demand is cut.
7. Write remediation before anyone fails. For each level, state the drill that fixes the common failure, the point the learner re-enters, and the new proof they retry. Remediation that amounts to doing the same task again slower is not remediation.
8. Validate the ladder. Save the ladder as JSON matching the field set in the examples and run bin/ladder_check.py <file>. Fix every reported problem. Then read the ladder once as a learner: at each level, could you tell what to do next and how you would know you had done it?
9. Place learners by proof, not by claim. A learner with prior experience takes the proof task for the level they claim, plus that level's foundation check. Passing both places them at the next level. Self report, years of exposure, and course completion never place anyone on their own. Progress runs on evidence of mastery, not seat time.

Output Contract

A finished ladder always contains:

• Topic, terminal performance, scope boundary, and prerequisites
• Exactly five levels in order, each with: name, at least two observable can do items, a not yet list, how support fades at that level, one proof task on unseen material, a pass standard, a gate check, and a remediation path
• A foundation check on every level from 2 to 5
• A placement rule for learners with prior experience
• A validation line showing bin/ladder_check.py passed on the JSON version

Operating Rules

1. Accuracy rule: every factual claim about a learning model must trace to references/research-brief.md. Never invent stages, dates, or findings. Unverifiable means cut it or label it as a design choice, as this skill does with its own level names.
2. Never advance a learner on time served, effort, confidence, or a single good performance on practised material. Proof on unseen material, plus the foundation check, is the only route up.
3. Never use a fixed hour count as a level requirement. Hours vary by learner and topic, and a famous round number for practice hours is not a standard any source in the brief supports.
4. Keep the scope boundary honest. When a learner asks about work outside the ladder, name the next ladder or say it is out of scope. Do not stretch level 5 to cover it.
5. When a proof fails, diagnose before prescribing: wrong level placement, a missing foundation item, or practice that did not match the proof. Each cause has a different fix, and the remediation path in the ladder names which one applies.
6. Read a weak proof with SOLO in mind before scoring it. Several correct disconnected points call for connection work. One connected point calls for breadth. The feedback differs, so the diagnosis comes first.