workflow-optimizer

When to Use

Optimize workflow — info access, token economy, caching, auto-learning triggers.

Core Principles

1. Info Access: right info, right time. 2. Token Economy: every token earns its place. 3. Auto-Learning: patterns→rules→habits→instinct. 4. Cache Aggressively: compute once, reuse everywhere.

Speed Patterns

1. Pre-flight: mem_search("<keywords>", limit=3) → top 3; check skill-graph; no recent error (immune-system). Savings: 2-5K/task.
2. Lazy loading: Q&A → karpathy-loop + lean-context (skip sdd-, judgment-day) | Bug fix → recovery-protocol + immune-system (skip sdd-propose) | Review → code-review-agent + judgment-day (skip sdd-). Savings: 5-15K/session.
3. Compression: L1 (8 msgs/15 calls) -60-70% | L2 (20 msgs/3 L1) decisions+ID -40-50% | L3 (ORANGE>60%) 1-liner+ref -80-90%.
4. Parallel gather: >3 files → delegate explore; independent research → parallel; batch reads → single call.

Auto-Learning Triggers

IF same_fix_2x → ANTI-PATTERN-CATALOG · IF same_error_3x → immune-system · IF user_correction_2x → AGENTS.md rules · IF repeat_workflow_3x → create skill. Pipeline: Capture (mem_save every decision/bugfix/discovery) → Extract (dreaming on 5th error) → Evaluate (score delta) → Apply (inject next session).

Information Access

Fast recall: mem_context → mem_search → mem_get_observation. Proactive: project → mem_search(project=...); feature → query; error → mem_search(type=bugfix).

Reference

Extended details, examples, patterns → docs/skills/workflow-optimizer/reference.md

Anti-Rationalization

Rationalization	Red Flag	Verification
"Caching no aplica aquí"	Recomputing instead of reusing mem_search	Check pre-flight mem_search cache — saves 2-5K/task if skipped
"Paralelizar es overkill"	Serial gather for >3 files	>3 files → delegate explore parallel — verify batch vs sequential cost
"Auto-learning después"	Deferring mem_save after decision/bugfix	Capture→Extract→Evaluate — mem_save every decision/bugfix now

Red Flags

• Doing work without checking output format → STOP, re-read skill
• Second occurrence of same rationalization → force RED zone

Verification

• Output matches skill ## Output contract + file:line citaton
• cross-ref-check.ps1 → SKILL.md OK

Refs

Cross-Refs: skill-graph | performance | context-watchdog | lean-context | karpathy-loop