skills.rest
OpenLAIR
OpenLAIROFFICIAL
Explore all skills by this author →
GitHub

ds-optimize

Manages candidate briefs, optimization frontiers, and branch promotion for algorithm-first research quests.

1.1k|116|Updated Feb 26, 2026
Search Tags:#Optimization#research workflow#candidate ranking#branch promotion#experiment routing#algorithm research#frontier management
One-click install
npx skills add https://github.com/OpenLAIR/dr-claw --skill ds-optimize
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ds-optimize
Source: https://github.com/OpenLAIR/dr-claw/tree/main/skills/ds-optimize
Command: npx skills add https://github.com/OpenLAIR/dr-claw --skill ds-optimize

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Algorithm-first research quests often drift into unfocused tinkering or premature paper writing. This Skill imposes a disciplined optimization control layer that turns loose ideas into ranked candidate briefs, promotes only the strongest directions into durable lines, and routes every measured result toward explore, exploit, fusion, debug, or stop decisions.

Core Features & Use Cases

  • Candidate Brief Management: Shapes fuzzy directions into structured candidate briefs with bottleneck, mechanism, expected gain, and risk fields before any branch is created.
  • Frontier-Driven Routing: Recovers the optimization frontier via artifact and memory calls, then selects exactly one dominant next action per pass across six internal submodes (brief, rank, seed, loop, fusion, debug).
  • Durable Record Keeping: Maintains OPTIMIZE_CHECKLIST.md and CANDIDATE_BOARD.md as quest-visible control surfaces, and records every candidate, result, and route decision through artifact and memory APIs.
  • Use Case: A researcher optimizing a model training pipeline uses this Skill to rank three candidate mechanisms, promote the strongest into a durable line with its own worktree, run bounded smoke checks, and record a fusion route when two lines show complementary strengths.

Quick Start

Ask the assistant to enter the optimize stage for the current algorithm-first quest and rank the existing candidate briefs against the current optimization frontier.

Frequently Asked Questions about ds-optimize

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I manage optimization candidates in an algorithm-first research quest?▼

Create branchless candidate briefs with artifact.submit_idea using submission_mode='candidate', rank them on one shared comparison surface, and promote only the top 1-3 into durable lines. Track everything in OPTIMIZE_CHECKLIST.md and CANDIDATE_BOARD.md.

What is the difference between a candidate brief and a durable optimization line?▼

A candidate brief is a branchless method-level proposal recorded with submission_mode='candidate'. A durable line is a promoted direction with submission_mode='line' that opens a real branch or worktree and becomes a formal optimization path.

When should I use the optimize stage instead of the experiment stage?▼

Use optimize when the quest is algorithm-first, the baseline gate is confirmed, and the main need is ranking directions or managing explore, exploit, fusion, and debug routes. Hand off to experiment once a promoted line needs implementation and measured runs.

How does the frontier decide between explore, exploit, fusion, and debug?▼

The frontier is read via artifact.get_optimization_frontier at the start of each pass. Explore fits when no line dominates, exploit when one line clearly leads, fusion when two lines have complementary strengths, and debug when a valuable candidate failed for a fixable reason.

When should I not use this optimization workflow?▼

Avoid it when the baseline gate is unresolved, when the main need is a paper draft, rebuttal, or review, or when the quest is still in broad literature scouting with no concrete optimization handle.

Related Skills

Explore More Education & Research Skills→
PantaniPantani
30

td-hype-synthesize

Synthesize scout outputs into a prioritized HYPE_TOOL_BACKLOG.md with vet notes.

Community
Advanced
brand6brand6

smoke-check

Run automated smoke tests and generate PASS/FAIL reports for QA readiness.

Community
Advanced
Euda1mon1aEuda1mon1a

recall

Search episodic task history for lessons from prior work.

Community
Intermediate
ResearAIResearAI
3.3k

write

Generate publication-ready paper drafts from validated quest evidence.

Community
Advanced