agent-sort
Build an evidence-backed ECC install plan for a specific repo by sorting skills, commands, rules, hooks, and extras into DAILY vs LIBRARY buckets using parallel repo-aware review passes. Use when ECC should be trimmed to what a project actually needs instead of loading the full bundle.
Security Assessment
About agent-sort
The `agent-sort` skill is designed to streamline the installation and configuration of ECC components for specific repositories. Rather than loading the full ECC bundle, which can be overly broad and noisy, this skill classifies skills, commands, rules, hooks, and additional components into two distinct buckets: DAILY and LIBRARY. DAILY components are essential elements that should be loaded every session for the repository, while LIBRARY components remain accessible but are not loaded by default. This approach ensures that the ECC environment is tailored to the actual needs of a project, with classification decisions grounded in concrete evidence from the codebase rather than assumptions or generic preferences.
Key features of `agent-sort` include evidence-backed classification using repository-local data such as file extensions, package managers, framework configurations, CI and hook setups, build and test scripts, dependency manifests, and documentation describing the stack. The skill supports parallel or sequential review passes across agents, skills, commands, rules, hooks, scripts, and other extras, enabling systematic assessment of all components. The workflow involves reading the repository to establish its language and framework stack, building an evidence table for each component, and then deciding which components belong to DAILY versus LIBRARY based on actual usage and relevance. The skill also generates outputs like DAILY and LIBRARY inventories, an install plan, verification reports, and optionally a `skill-library` router for easier navigation of the components.
This skill is particularly useful for development teams that want a repeatable, reproducible process for trimming ECC installations to what a project actually requires. It is ideal when the full ECC bundle is unnecessarily large, when the repository stack is clear but manual curation would be tedious, or when a project has accumulated irrelevant languages, hooks, or rules over time. Target users include software engineers, DevOps teams, and project maintainers seeking a clean, evidence-driven, and automated method to optimize their ECC setup for efficiency and accuracy.
FAQ
When should I use the `agent-sort` skill?
Use it when you want a project-specific ECC setup rather than the full bundle, especially if the project only uses a subset of ECC components or if the repository has drifted into unnecessary languages or hooks.
What types of outputs does `agent-sort` produce?
It produces DAILY and LIBRARY inventories, an install plan, a verification report, and optionally a `skill-library` router for easier access to LIBRARY components.
How does `agent-sort` determine whether a component is DAILY or LIBRARY?
Classification is based on concrete evidence from the repository, such as file types, package managers, framework configurations, CI setups, scripts, and documentation. DAILY components are essential for every session, while LIBRARY components are kept accessible without being loaded by default.
Does `agent-sort` require subagents to run in parallel?
No, parallel subagents can speed up the review process, but the same classification passes can be performed sequentially if subagents are not available.
Are there any limitations or non-negotiable rules?
Yes, classification must rely on the current repository as the source of truth, every DAILY decision must cite concrete evidence, hooks or rules that the repo cannot use should not be installed, and ECC-native surfaces are preferred without introducing secondary install systems.
Install agent-sort
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
affaan-m/everything-claude-code