Guide · Agent skill
How to use the Agent Sort agent skill (ECC install trimmer)
Updated · written with AI from the project's own documentation and checked against our catalogue data
Agent Sort is an agent skill from the affaan-m/ECC repository that helps you install only the parts of the ECC toolkit your project needs. Instead of loading the full bundle, the agent inspects your repository and sorts skills, commands, rules, hooks and extras into two buckets. This guide explains what the skill does, how to install it, how to ask for it, and what to check afterwards.
Agent Sort (ECC install trimmer)affaan-m/ECC · MIT Download the skill Quick start
- 1Get the ZIP and unzip it. Download the agent-sort ZIP from this page and unzip it. The skill is a folder containing SKILL.md.
- 2Place the skill. Put the folder in ~/.claude/skills/ for personal use or in your project's .claude/skills/ folder. Alternatively, upload the ZIP in the Claude apps under Settings → Capabilities → Skills.
- 3Make sure ECC is installed. The skill classifies an existing ECC installation from the affaan-m/ECC repository, so ECC needs to be available to sort.
- 4Open your target repository. Start the agent inside the repository you want to trim, so it can read its files.
- 5Ask for a trimmed install plan. Request an evidence-based DAILY and LIBRARY split for this repository in plain language.
- 6Review the evidence table. Check each DAILY decision against the cited files, and treat any recommended removals as items needing your confirmation.
- 7Check the verification report. Confirm the DAILY and LIBRARY counts, the removed old surfaces, and the open questions before relying on the new setup.
What is an agent skill, and what does agent-sort do?
An agent skill is a folder containing a SKILL.md file with instructions that an AI agent can load when a task matches the skill's description. The agent-sort skill covers one specific task: deciding which ECC components a particular repository should use.
When a project only needs part of the ECC toolkit, the skill makes the agent inspect the repository, including file types, package manager files, CI and hook configs, and test and build scripts. It then sorts every ECC component into one of two buckets: DAILY, which is loaded every session, and LIBRARY, which stays accessible but is not loaded by default.
- DAILY: components that match your repository's languages, frameworks, workflows or operator surfaces, and should load in every session.
- LIBRARY: components that are useful to keep but not worth loading by default. They stay searchable or reachable through a router skill or manual use.
- Output: an install plan with the evidence for each decision, plus a verification report.
Install the agent-sort skill with a ZIP file
Download the ZIP for the skill from this page and unzip it. The skill ships as a single file, SKILL.md. Place the unzipped skill folder in one of these locations:
- Personal skills: ~/.claude/skills/ (available across your projects).
- Project skills: .claude/skills/ inside the repository you want to trim (available only in that project).
- Claude apps: upload the ZIP under Settings → Capabilities → Skills.
Ask for the skill with a natural request
You do not need a special command. Describe the outcome you want in plain language, and mention that you want only the parts of ECC your repository uses. The skill's description tells the agent when to use it, so requests about trimming, leaner setups, or cleaning up drifted language rules are the right kind of trigger.
The source notes the skill is written in Japanese, and its own description is in Japanese. Plain English requests should still work, but you can phrase requests in either language.
- Trim my ECC setup to only what this repository needs.
- Which ECC skills, rules and hooks should I keep for this project, and which should go to the library?
- Our install has Python rules but this repo has no Python. Build an evidence-based ECC install plan.
- Sort the ECC components for this repo into DAILY and LIBRARY, and show the evidence for each one.
What the agent checks in your repository
The skill asks the agent to establish the real stack before classifying anything. It records the languages and frameworks in use, the main package manager, the test stack, the lint and format stack, deployment or runtime surfaces, and any operator integrations already present.
The skill's evidence sources include file extensions, package manager files and lockfiles, framework configuration, CI and hook configuration, build and test scripts, imports and dependency manifests, and repository documentation. The source gives these example commands, which use rg (ripgrep) and cat:
rg --files
rg -n "typescript|react|next|supabase|django|spring|flutter|swift"
cat package.json
cat pyproject.toml
cat Cargo.toml
cat pubspec.yaml
cat go.modRead the evidence table and install plan
For each candidate component, the skill builds a table row with the component path, its type, the proposed bucket, the repository evidence, and a short reason. The result is easy to review: every DAILY decision should cite concrete evidence from your repository.
The skill's rules say LIBRARY means keep accessible, not delete. However, the install plan can still recommend removals, so read those lines with care and confirm them yourself before any file is deleted. The final output is returned in this order: STACK, DAILY, LIBRARY, INSTALL PLAN, and VERIFICATION.
- DAILY skills go into .claude/skills/, or stay there if already present.
- DAILY rules: only the rule set matching your language is installed.
- DAILY hooks and scripts: only those compatible with your environment are kept.
- LIBRARY items: kept accessible through search or the optional skill-library router.
skills/frontend-patterns | skill | DAILY | 84 .tsx files, next.config.ts present | コアフロントエンドスタック
skills/django-patterns | skill | LIBRARY | no .py files, no pyproject.toml | このリポジトリではアクティブでない
rules/typescript/* | rules | DAILY | package.json + tsconfig.json | アクティブな TS リポジトリ
rules/python/* | rules | LIBRARY | zero Python source files | アクセス可能に保つのみUse parallel review when your runtime supports it
If your agent runtime supports subagents, the skill splits the review into parallel passes: agents, skills, commands, rules, hooks and scripts, and extras. If subagents are not available, the same passes run one after another. This is optional and does not change the result format.
Verify the result and hand off next steps
After the plan is applied, the skill asks the agent to check that every DAILY file is in its expected place, that outdated language rules are no longer active, that no incompatible hooks are installed, and that the result matches the repository stack. The report lists the DAILY and LIBRARY counts, any old surfaces removed, and open questions.
If the next step is interactive installation or repair, the skill hands off to configure-ecc. For duplicate cleanup or catalogue review, it hands off to skill-stocktake. For trimming broader context, it hands off to strategic-compact.
What people use it for
- Keeping only TypeScript or React rules and skills in a front-end repository that has no Python code.
- Cleaning up an ECC setup where Python or Java rules were loaded into a project that does not use those languages.
- Auditing which hooks and scripts are compatible with your current operating system and runtime before keeping them.
- Creating an optional skill-library router so rarely used ECC references stay searchable without loading every session.
Frequently asked questions
Is the agent-sort skill free to use?
The skill is released under the MIT licence, as stated on this page. The source does not mention pricing for it, and it requires an existing ECC installation from the affaan-m/ECC repository.
What do I need before I can use agent-sort?
You need an ECC installation from the affaan-m/ECC repository to classify, and access to the target repository's files. The skill's example commands use rg (ripgrep) and cat. Optional subagent support in your agent runtime lets the review run in parallel.
Does agent-sort delete files from my project?
The skill's rules say LIBRARY means keep accessible, not delete. However, the install plan may still recommend removals, which need human review. Confirm any removal before it happens.
Where do I put the agent-sort skill folder?
Place it in ~/.claude/skills/ for personal use across projects, or in .claude/skills/ inside a specific project. In the Claude apps you can upload the ZIP under Settings → Capabilities → Skills.
Does agent-sort run scripts or use the network?
According to this page, the skill runs no scripts and uses no network. It works by having the agent read and search your repository files.
Is agent-sort a good fit for every repository?
No. If your repository already works well with the full ECC bundle, the process adds effort without clear benefit. It is also not suited to non-developers or anyone without access to the repository files, since it depends on inspecting the codebase.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Agent Sort is worth using if your repository uses only a few languages or frameworks and you want a leaner, evidence-backed ECC setup. Skip it if the full bundle already works for you, or if you cannot give the agent access to the repository files. Review every recommended removal yourself, because the skill keeps LIBRARY items accessible but its plan can still suggest deletions.
Download the skill