This repository is written for agents and automated consumers, not for tutorials, marketing, manual operation, or one specific AI runtime.
Reader comfort is secondary. If prose polish conflicts with agent execution clarity, keep the agent contract.
The packages are runtime-agnostic agent method packages. SKILL.md is the portable agent-readable contract; runtime-specific metadata belongs in adapters/ only after evidence-gated validation.
Aha names the discovery moment when an agent identifies the skill or capability that matches the target need. Source-surface expansion, query expansion, surprising-but-traceable hits, and raw signals are methods for reaching that moment; adoption judgment stays in later gates.
Timing is part of fit: search when the ecosystem, adoption window, and actual need line up.
| Path | Purpose | Status |
|---|---|---|
aha-skills-finder/ |
Surface skill/capability candidates a first-pass search would likely miss, with traceable sources and raw signals. | v0 draft |
License: MIT.
aha-skills-finder exists for the moment when an agent must identify the skill
or capability that matches the target need, not merely the first candidate an
obvious query returns. It widens the source surface, keeps plausible surprises
traceable, and records why each candidate appeared before any later audit or
adoption stage judges it.
It covers gaps that third-party skill libraries, installers, routers, marketplaces, and registries usually do not cover by themselves:
- broad cross-surface discovery instead of one registry, catalog, or package ecosystem;
- lane-first decomposition before searching mixed outcomes;
- exploratory query expansion from seed terms into ecosystem synonyms, tooling verbs, runtime terms, package terms, claim terms, local-language terms, and spelling variants;
- multiple traceable candidates instead of one recommendation;
- raw signals only for stars, downloads, installs, registry stats, package metadata, README claims, and list labels;
- source gaps and lane-relative false positives instead of premature no-hit or global rejection claims;
- agent-readable handoff artifacts for later audit or adoption stages.
It does not recommend adoption, rank winners, audit safety/source quality, install, enable, configure, mutate runtime config, log in, publish, deploy, or perform browser/provider/credential/live/external-write actions.
This repository is for runtime-agnostic skill method packages and machine-readable example artifacts. Keep package docs short and agent-entry oriented:
- when to use the skill;
- required inputs;
- required outputs;
- schemas and light red-line validators;
- hard non-goals;
- examples as canonical artifacts.
Canonical examples are limited to:
aha-skills-finder/examples/find-skill-finder/aha-skills-finder/examples/find-skill-audit/
Do not turn package docs into tutorials, marketing pages, adoption audits, install managers, or hosted search products.
Do not add quickstarts, walkthroughs, screenshots, comparison prose, landing pages, or docs-site structure unless they directly improve agent execution.