feat: add optimize-sorting Python workflow with gap detection and I/O profiling

[colehurwitz]

Changes

• New file: .factory/workflows/optimize_sorting.py — Project-local Python workflow definition encoding the full optimize-sorting workflow graph (31 nodes, 39 edges) with two new capabilities:

  • Gap Detection: Enhanced researcher_profile_pipeline prompt with gap analysis that compares speed_seconds.mean to the sum of stage means, using a 10% threshold, and outputs a gap_analysis JSON section identifying unexplained overhead
  • I/O Profiling: Modified FnNode commands for lock_baseline, run_benchmark_t*, and confirm_benchmark_t* to capture /proc/self/io before and after benchmark runs, computing read/write byte deltas with graceful fallback (available: false) when procfs is unavailable

• New file: tests/test_optimize_sorting_workflow.py — 61 tests covering:

  • Workflow loads correctly with expected node/edge counts
  • All 31 nodes exist with correct types (8 AgentNode, 11 FnNode, 12 GateNode)
  • Gate nodes have evaluator commands; agent nodes have post_checks
  • Gap detection keywords present in researcher prompt
  • I/O capture logic present in benchmark scripts with graceful fallback
  • Archivist nodes use haiku model with 300s timeout
  • Graph validation passes with no structural issues
  • Gap arithmetic threshold tests

[github-actions]

Sentrux Quality Report

Absolute

Scanning ....
[scan] git ls-files: 631 total, 617 kept, 14 dropped (ext:14, meta:0, big:0)
[build_project_map] 617 files, 112 unique dirs, 103 cache misses, 5.4ms
[resolve] 1162 resolved, 1575 unresolved (of 2737 total specs)
[resolve_imports] project_map 5.6ms, suffix_idx 1.1ms, suffix_resolve 19.3ms, total 25.9ms
[build_graphs] 617 files | maps 2.3ms, imports 26.1ms, calls+inherit 6.5ms, total 34.9ms | 1161 import, 8388 call, 1 inherit edges
sentrux check — 3 rules checked

Quality: 4447

✗ [Error] max_cc: 6 function(s) exceed max cyclomatic complexity of 30
    factory/cli/_ceo_helpers.py:_validate_ceo_flags (cc=43)
    factory/cli/_ceo_helpers.py:_execute_ceo (cc=43)
    examples/chess_evolve.py:main (cc=33)
    factory/cli/_task_builder.py:_build_ceo_task (cc=32)
    factory/cli/run.py:cmd_run (cc=32)
    factory/outer_loop/mutations.py:validate_and_repair (cc=31)

✗ 1 violation(s) found

Diff (vs base branch)

Scanning ....
[scan] git ls-files: 631 total, 617 kept, 14 dropped (ext:14, meta:0, big:0)
[build_project_map] 617 files, 112 unique dirs, 103 cache misses, 5.8ms
[resolve] 1162 resolved, 1575 unresolved (of 2737 total specs)
[resolve_imports] project_map 5.9ms, suffix_idx 1.3ms, suffix_resolve 18.8ms, total 26.0ms
[build_graphs] 617 files | maps 2.3ms, imports 26.2ms, calls+inherit 6.8ms, total 35.4ms | 1161 import, 8388 call, 1 inherit edges
sentrux gate — structural regression check

Quality:      4447 -> 4447
Coupling:     0.80 → 0.80
Cycles:       4 → 4
God files:    3 → 3

Distance from Main Sequence: 0.35

✓ No degradation detected

[colehurwitz]

SKILL.md Regeneration Fix

Added the regenerated workflow-optimize-sorting/SKILL.md and SKILL.annotations.yaml — these were missing from the original commit.

Before: skills cache SKILL.md (59KB, 841 lines) — no gap detection or I/O profiling content
After: regenerated SKILL.md (104KB, 2521 lines) — includes Gap Detection (1 match) and capture_io (42 matches)

Also reverted unrelated workflow-design-v2/ SKILL changes and refreshed the local skills/ cache.

[colehurwitz]

Superseded by unified PR — see #1443