loop-discovery

Routes asset-creation requests through evidence, coverage, and carrier checks before persisting reusable assets.

2|Updated Jun 8, 2026
One-click install
npx skills add https://github.com/HACK-WU/skills --skill loop-discovery-hack-wu
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: loop-discovery
Source: https://github.com/HACK-WU/skills/tree/main/skills/loop-discovery
Command: npx skills add https://github.com/HACK-WU/skills --skill loop-discovery-hack-wu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Teams and AI agents tend to create a new skill, rule, or memory for every request, causing asset sprawl, trigger conflicts, and unmaintainable knowledge bases. This Skill acts as a gatekeeper that decides whether an asset should be created at all before any creation happens. ## Core Features & Use Cases - Evidence Gate: Checks whether the request is backed by repeated occurrences (≥2 comparable tasks), high-cost single events, or explicit future requirements before allowing creation. - Coverage Ladder: Scans existing skills, rules, solutions, and memories in five ordered tiers to decide between reusing, fixing triggers, extending an existing asset, or creating new. - Carrier Selection: Chooses the minimal appropriate carrier (memory, rule, solution, skill, script, or expert asset pack) based on content shape. - Use Case: After solving a problem, the AI wants to call solution-capture. This Skill first verifies the evidence, finds an existing skill already covers the scenario but was never triggered, and concludes with a "fix trigger" action instead of a duplicate creation. ## Quick Start Ask the AI to run a loop-discovery routing check on whether a proposed skill, rule, or memory should be created before persisting it.

Frequently Asked Questions about loop-discovery

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

FAQPage Schema
How do I decide whether to create a new AI skill or reuse an existing one?

Run a routing check that first validates evidence (repeated occurrences or high cost), then walks a coverage ladder over existing skills, rules, and memories. If an existing asset covers the scenario but was not triggered, fix its description instead of creating a duplicate.

What is the difference between a memory, rule, solution, and skill?

A memory stores one-line facts or preferences, a rule is an unconditional constraint applied every conversation, a solution documents steps for one specific problem, and a skill is a multi-step workflow with trigger conditions. Choose the lightest carrier that fits the content.

When should repeated work not be turned into a reusable asset?

Single low-cost occurrences and speculative future needs should not be persisted. If the platform already handles the task natively, or the evidence is only the AI's own judgment without concrete task history, defer and leave an observation point.

How do I prevent skill sprawl and trigger conflicts in an AI agent setup?

Gate every creation request through an evidence check and a coverage scan of existing asset descriptions before creating anything. Prefer fixing trigger phrases or extending existing assets over new files, and route confirmed creations to dedicated creator skills.

Can a user override the routing conclusion and force asset creation?

Yes. The final decision belongs to the user. If the user insists after discussion, the skill provides a risk warning about thin content or low trigger value and then proceeds with creation as requested.