ALGO-01 — Shared algorithms & optimization capability: F# library + product skill knowledge pack, available in product workspaces #5222

Work level: Outcome · Track: Active delivery

Promise

A growing, shared algorithms and optimization capability from classic AI and optimization research, available both in the framework and in every product workspace, starting with games scaffolded from fs-gg-fable-game. It covers:

• pathfinding (A*, JPS, HPA*, flow fields, D* Lite);
• decision making (behaviour trees, HTN, GOAP, utility AI, FSMs);
• adversarial and stochastic search (minimax and alpha-beta, MCTS);
• spatial structures (influence maps, spatial hashing, k-d trees);
• optimization (LP and ILP, local search, simulated annealing, GA, constraint solving).

It has three connected parts:

1. An F# library on nuget.org. Tested implementations, admitted only with property tests (for example A* optimality against BFS) and bench numbers from the FS.GG.Game harness (the corpus case 41-residue-and-full-queue is FLAKY in CI — it exits 5 at random, and a gate that fails randomly teaches people to re-run it #761/fix(engine): next/take name the observed reason, not a guessed list of causes (#440) (case 41) #792). Fable-portable where useful.
2. A product skill, fs-gg-algorithms, with a knowledge pack. One cited entry per technique: idea, complexity, when and when not to use it, pitfalls, F# usage, references (summarised, never copied), and bench numbers. It reuses the Atlas record format and search tooling as a product-local archive, so it's searchable inside a game workspace.
   • Distribution: through the existing product-skill route (the producer's template/skill-manifest/skill-manifest.json, registry/skills.yml, ADR-0017 materialization into scaffolds, and re-seeding via fsgg-sdd upgrade).
   • Existing skills: it becomes the umbrella over the existing algorithm skills (fs-gg-grids, fs-gg-collision, fs-gg-line-drawing, fs-gg-visibility) rather than duplicating them.
3. A growth loop. Framework Deliveries that implement or tune an algorithm add or update an entry at closure, and product workspaces have a documented route to contribute entries upstream.

The framework Atlas holds only the decisions about this capability, as pointers. It is not the knowledge pack.

Source: the owner's request on 2026-11-01. The pack must be available in product workspaces, not only in the framework Atlas.

Acceptance

• A freshly scaffolded fs-gg-fable-game workspace contains the skill and a searchable pack, and can reference the library.
• At least 15 entries are seeded with tested implementations harvested from existing reviewed code (RuneScape A* with doors, the behaviour and arbitration model, nearest queries, the grids, collision and visibility skills).
• An existing game workspace receives updates via fsgg-sdd upgrade.
• The capture rule is active.