| name | pathfinder-review |
| description | Multi-agent Zig review of the pathfinder module for coherency, cohesion, and standards adherence. Use when the user asks for pathfinder review, pathfinding module review, /pathfinder-review, or a module-wide pass on src/game/systems/pathfinding/. |
Pathfinder Review
Orchestrate a multi-phase review of the pathfinder module. You (parent) spawn
all subagents — children cannot nest. Do not edit code unless the user asks for
fixes after the report.
Agents: zig-review-specialist (.grok/agents/).
Unit table: .grok/skills/zig-workflows/references/module-presets.md → pathfinder.
Standards preamble (include in every unit prompt)
Before reviewing, read:
- docs/coding-standards.md
- docs/simulation-tiers-and-pipeline.md
- docs/architecture.md
Hot/frame paths must be allocation-free after init/reserve/warmup; no per-frame
string lookups, hash-map dispatch, broad dynamic dispatch, or formatted logging
on hot paths. Production APIs expose runtime concepts only.
Lenses (every unit)
- COHERENCY — consistent invariants, units/coordinate spaces, edge cases
- COHESION — single responsibility, ownership boundaries, duplication
- STANDARDS — Zig style, hot-path discipline, explicit error sets, tests
Phase 1 — Per-unit review (parallel)
Spawn one zig-review-specialist per preset unit (system, nav_graph,
caches, solve, nav_grid, types_and_memory, group_field_and_scratch,
facade_and_test_support, benchmarks).
subagent_type: zig-review-specialist
capability_mode: read-only
background: true
description: pathfinder review <unit>
Prompt template:
Review unit "<unit>" of the ZeroLight-Framework pathfinder module.
Files (read in full):
- <files from preset>
Context: <note from preset>
<standards preamble>
<lenses>
Return severity-ordered findings (critical → nit). Each finding: severity,
category (coherency|cohesion|standards|performance|correctness), file, line,
title, detail, suggestion. If clean on a lens, say so in the summary — do not pad.
Collect all results with get_command_or_subagent_output before Phase 2.
Phase 2 — Cross-cut (parallel)
Three more zig-review-specialist spawns (background: true):
| Unit | Focus |
|---|
module-coherency | Cross-file invariants; request lifecycle system → solve/nav_graph/nav_grid/caches |
module-cohesion | File split, facade surface, boundary violations, dead/duplicated code |
standards-and-hotpath | Post-warmup alloc, hot-path logging, naming, co-located tests |
List all pathfinder + benchmark files from the preset in each prompt.
Phase 3 — Synthesize (parent or one specialist)
Either synthesize yourself or spawn one zig-review-specialist with the full
corpus. Output one Markdown report:
- Executive summary (health across three lenses)
- Severity-ordered findings (merge duplicates; drop spurious; note corroboration)
- Each finding: severity, category,
file:line, problem, concrete fix
- Cross-cutting themes
- Prioritized action list
Do not invent findings beyond the corpus. Do not fix code unless asked.
Done criteria
- All Phase 1 units completed (or explicitly noted as failed/skipped)
- Cross-cuts completed
- Single severity-ordered report delivered to the user
