Define the interview skill graph and mastery evidence policy

[rosshd]

Parent: #56

Depends on: #45, #46, #48, and the interview-profile child of #56.

Problem

The shipped algorithms template is a useful course outline, but it is not a canonical interview-readiness model.

It does not define stable skill identities, prerequisites, practice evidence, transfer requirements, or the difference between knowing a data structure and independently recognizing when to use it.

Without that structure, adaptive selection and readiness reporting will be driven by model prose rather than durable rules.

Goal

Define a versioned interview skill and pattern graph that supports curriculum generation, adaptive problem selection, mastery decisions, and learner-visible readiness.

Initial skill domains

• Complexity analysis and constraint reading.
• Arrays, strings, hashing, linked structures, stacks, and queues.
• Two pointers, sliding window, prefix or suffix aggregation, and intervals.
• Binary search and search-space reduction.
• Trees, heaps, tries, graphs, BFS, and DFS.
• Recursion, backtracking, greedy reasoning, and dynamic programming.
• Union-find, topological ordering, and shortest paths where appropriate to the target level.
• Edge-case design, testing, debugging, and counterexample construction.
• Problem clarification, plan communication, complexity explanation, and follow-up adaptation.

Graph contract

• Use stable human-readable skill IDs.
• Separate concepts, patterns, process skills, and communication skills while allowing typed prerequisite edges.
• Define minimum evidence types for recognition, explanation, production, transfer, and delayed retrieval.
• Define which skills can be inferred from a problem attempt and which require an explicit check.
• Allow a problem to exercise multiple skills without awarding all of them automatically.
• Version graph content and mastery rules.
• Preserve historical evidence when graph definitions change.
• Keep learner-specific mastery outside the static graph.

Mastery policy

• Pattern mastery requires unaided production or novel transfer, not reading an editorial or passing one familiar problem.
• Hint use, partial code, copied structure, and worked examples reduce or defer mastery credit.
• Complexity and edge-case claims require explicit evidence.
• Communication evidence is recorded separately from algorithm correctness.
• Repeated success on near-duplicate problems must not count as broad transfer.
• Blocking prerequisites affect selection but must not trap the learner in endless drilling.

Acceptance criteria

• The graph has stable IDs, typed prerequisites, version metadata, and validation.
• Every initial skill declares qualifying evidence and transfer expectations.
• Problems can reference multiple skills with primary and supporting roles.
• Adaptive selection can query ready, blocked, weak, due, and unassessed skills deterministically.
• The learner can inspect why a skill is considered weak, provisional, or ready.
• Graph changes do not destroy or silently rewrite historical evidence.
• Tests reject cycles or invalid edges where the contract forbids them.
• Fixtures cover an advanced learner, a prerequisite gap, hint-dependent success, and delayed transfer failure.
• The algorithms template can seed an interview course from the graph without becoming the canonical data source.
• make check passes.

Out of scope

• Exhaustive competitive-programming coverage.
• Company-specific frequency claims without an authorized data source.
• System design and behavioral interview skills in the first version.
• Model-generated graph mutations during normal tutoring.

Documentation

• Document canonical terminology in CONCEPTS.md if the file exists.
• Document mastery implications in docs/TUTOR_INTERACTION.md and docs/LEARNING_SCIENCE.md.

[rosshd]

Board hygiene status: implementation is complete and validated on local main at 50e468f, but that commit is not yet present on origin/main. Leaving this issue open until the local main branch is pushed; close it once the implementation is visible remotely.