algorithm-tutorial-build

name	algorithm-tutorial-build
description	v0.1.0 - Build exactly one planned algorithm tutorial checkpoint and stop for review. Use when a task-first algorithm tutorial plan is ready and the next section must derive one mechanism, invariant, code fragment, or runnable checkpoint from concrete pressure without self-approval.

Algorithm Tutorial Build

Overview

Draft one planned algorithm tutorial checkpoint and stop. This skill writes the next piece of the article, then hands it to algorithm-tutorial-review or algorithm-tutorial-deepen.

Build may run self-checks, but self-checks are evidence, not acceptance.

When to Use

• An algorithm-tutorial-plan task list exists.
• The next planned checkpoint should be drafted.
• The tutorial is about a method itself, not one OJ problem.

When NOT to use: planning, reviewing, deepening, simplifying, LeetCode problem tutorials, or writing a whole multi-section article in one pass.

Reference Map

• ../references/derivation-first-explanations.md
• ../references/depth-checklist.md

The Build Loop

1. Load One Task
   • Read dependencies, acceptance criteria, verification, document target, code role, review gate, and deepening gate.
   • Verify: exactly one task is in scope.
2. Re-state Pressure and Baseline
   • Show the tiny task, trace, bottleneck, or missing capability.
   • Identify what the reader already has.
3. Draft One Mechanism
   • Add one representation, invariant, rule, helper, formula, or code block.
   • Introduce terms at first real use.
   • Keep code connected to the current checkpoint.
4. Add Evidence
   • Include a worked trace, invariant check, formula with variables, runnable snippet, or counterexample as required by the task.
   • Execute runnable checks when feasible.
5. Freeze and Stop
   • State what the checkpoint now explains or runs.
   • State what still lacks.
   • Output needs_review or needs_deepening.
   • Do not continue to the next task.

Output Format

## Build Result
- status: needs_review | needs_deepening
- task:
- document_target:
- code_change_role:

## Drafted Content
<one checkpoint only>

## Self-Checks
- ...

## Gate Handoff
- review_skill: algorithm-tutorial-review
- deepen_skill: algorithm-tutorial-deepen | n/a
- next_step_blocked_until: gate_pass

Common Rationalizations

Rationalization	Reality
"I can finish the whole article faster."	Hidden jumps compound across algorithm tutorials.
"The invariant is obvious."	The invariant is one of the teaching objects.
"The code works, so depth is optional."	Correctness and complexity are part of algorithm learning.
"Self-check passed."	Self-check is evidence; review or deepen gate accepts.

Verification

• Exactly one task was drafted.
• Concrete pressure appears before the mechanism.
• One mechanism or rule was introduced.
• Evidence is concrete.
• Output stops with a gate handoff.

Guardrails

• Do not self-approve.
• Do not continue to the next checkpoint.
• Do not write LeetCode-style problem solution flow for concept tutorials.
• Do not add detached final code.