manim-dsa-single-path

Manim DSA Single-Path Walkthrough

Relationship to the other two skills

Read manim-explainer-animations/SKILL.md first (owns the mobject/animation API surface), then manim-dsa-storytelling/SKILL.md for the full 3D-solid vocabulary table and the code-duality rules (real source in a Code mobject, synced SurroundingRectangle highlight, add_fixed_in_frame_mobjects for HUD text) -- this file only overrides the narrative shape, since a single-path video has one walkthrough instead of a brute-vs-optimized comparison. Reuse the solid vocabulary and code-duality rules verbatim; do not reinvent them here.

When to use this instead of manim-dsa-storytelling

Use this skill when the target is one of:

• A sort, traversal, or search with one standard approach (bubble/merge/quick/heap sort, BFS, DFS, Dijkstra, topological sort, MST algorithms)
• A tree or graph operation (max depth, LCA, diameter, cycle detection, bipartite check)
• A single-pass or single-table DP (fibonacci, climb stairs, knapsack, LCS) where there isn't a genuinely worse baseline worth animating
• A data structure's core mechanics (linked list reversal, stack/queue built from arrays, trie, segment tree, union-find)
• A bit trick or simple math routine (XOR tricks, sieve of Eratosthenes, GCD, fast exponentiation)

If a genuine brute-force baseline exists AND the complexity gap is real and instructive, use manim-dsa-storytelling instead -- don't force a two-path video into this one-path skill just because it's less work, and don't force a single-path video into a fake "brute force" frame just to reuse the comparison skill.

The narrative shape (4 acts, ~30-50s total)

1. Cold open -- the structure, spatially. Build the data structure/space with the solid from the vocabulary table (same table as manim-dsa-storytelling) before any code appears.
2. The walkthrough -- code and motion together. Bring in the real source in a Code mobject (real, verbatim) with a synced highlight tracking the active line, while a particle/marker moves through the 3D structure performing the algorithm. This is the largest act. For anything beyond ~8-10 elements/nodes/steps, subsample or use ChangeSpeed/run_time compression.
3. The key property. A short beat naming why this approach works -- loop invariant, structural guarantee, base case. Give it a real breath (Write + wait).
4. Payoff -- the result, held. Show the final state clearly (sorted array, shortest path traced in the accent color, returned value, built structure). Optionally caption the complexity as a fixed-in-frame label.