pollinationspollinations/search.elixpo/skills/optimize-search-runtime/SKILL.md

optimize-search-runtime

Skill24 starsChanged 7 months agoCC0-1.0View on GitHub

Select and execute the smallest sufficient OreoLook search plan. Use whenever a request may need current web evidence, source fetching, or deep research and runtime cost, latency, context size, tool count, and output depth must remain bounded.

In this repository13 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- communicate-naturallySkill
- export-documentsSkill
- handle-mediaSkill
- know-oreolook-creatorSkill
- make-imagesSkill
- orchestrate-searchSkill
- oreolook-personaSkill
- recall-memorySkill
- research-webSkill
- synthesize-answerSkill
- write-codeSkill
- write-contentSkill

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 5 are distinctive, and 3 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
Use quick for one live fact or a narrow lookup: at most 1 search, 2 fetched sources, and 350 answer tokens.
line 12
Very specific:
Use standard for one subject requiring verification or comparison: at most 2 searches, 4 fetched sources, and 700 answer tokens.
line 13
Very specific:
Use deep only for a genuinely multi-part investigation: at most 4 subqueries, 2 iterations per subquery, 4 fetched sources per subquery, …
line 14
Distinctive:
Stop as soon as the evidence supports the answer.
line 16
Distinctive:
Prefer the durable summary over raw older turns.
line 21
Distinctive:
Keep ordinary injected history within 3,000 tokens and deep history within 6,000 tokens.
line 22
Distinctive:
Keep retrieved semantic context within 4,000 characters and the five most relevant memories.
line 23
Distinctive:
Avoid repeating tool evidence or narrating internal work.
line 28

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

Optimize Search Runtime

Choose depth from meaning and required evidence, never keyword matching.

Plan

Use quick for one live fact or a narrow lookup: at most 1 search, 2 fetched sources, and 350 answer tokens.
Use standard for one subject requiring verification or comparison: at most 2 searches, 4 fetched sources, and 700 answer tokens.
Use deep only for a genuinely multi-part investigation: at most 4 subqueries, 2 iterations per subquery, 4 fetched sources per subquery, and 1,800 final tokens.
Reuse cached evidence and conversation summaries before requesting new work.
Stop as soon as the evidence supports the answer. Never spend remaining budget merely because it exists.

Context

Include only history needed to resolve references or maintain continuity.
Prefer the durable summary over raw older turns.
Keep ordinary injected history within 3,000 tokens and deep history within 6,000 tokens.
Keep retrieved semantic context within 4,000 characters and the five most relevant memories.

Output

Answer first and cite only sources actually used.
Avoid repeating tool evidence or narrating internal work.
Match response length to the request; a quick lookup should remain quick.

Runtime contract

agent: policy
tools: []
timeout_seconds: 1
max_concurrency: 1
output: bounded_search_policy

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free