Fallow Skill for Claude Code: What It Does and How to Install It
Makes your assistant run fallow on a TypeScript or JavaScript project to find unused code, duplicates, circular imports and risky changes.
- Installs
- 9.9K
- on skills.sh
- Stars
- 137
- on GitHub
- Files
- 8
- besides SKILL.md
- Updated
- Nov 2026
- last repo commit
SKILL.md tells the assistant to run fallow commands with --format json --quiet and read the result. It routes by intent, for example fallow audit --base <ref> when you commit or open a PR and fallow dead-code --trace before deleting an unused export.
It loads when you ask to audit PR risk, find unused code, detect duplicates or run fallow. It needs the fallow CLI in the project. The skill folder also holds 7 reference files and an agents/openai.yaml.
Good for
- Developers cleaning up unused exports, files and dependencies
- Reviewers who want a pass, warn or fail check on a pull request
- Teams ranking complexity hotspots and refactoring targets
Install fallow
Claude Code
- 1
Copy this prompt.
Install the fallow skill in Claude Code. Run this command: /plugin marketplace add fallow-rs/fallow-skills /plugin install fallow@fallow-skills Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Requires Node.js, fallow installed (npm install -g fallow).
Pick one route. The plugin routes install the fallow plugin: 5 skills (fallow, fallow-review, fallow-cloud-setup, fallow-setup, fallow-codex-onboarding), a Claude Code statusline command and a commit and push gate hook. In Codex the plugin also adds the Fallow app. The skills CLI and folder routes install the skill without the hook. All routes need fallow in the project: npm install -g fallow, npx fallow, or cargo install fallow-cli. There is no Claude app upload route because the skill runs a CLI.
From the project README. Source.
What's in the skill
When it loads
Codebase intelligence for TypeScript and JavaScript. Static analysis reports changed-code risk, cleanup opportunities, duplication, circular dependencies, complexity hotspots, architecture boundaries, design-system drift, feature flags, and opt-in security candidates. Optional local similar-code discovery finds functions that may implement the same intent despite different syntax. Runtime coverage can merge production execution data. Use when asked to audit PR risk, find unused code or dependencies, compare semantically similar functions, detect duplicates, inspect architecture boundaries, merge runtime coverage, auto-fix supported issues, or run fallow.
Sections in SKILL.md
- When to Use
- When NOT to Use
- Prerequisites
- Agent Rules
- Onboarding And Insight
- Task Cheat Sheet
- Commands
- Issue Types
- MCP server
- References
- Common Workflows
- Audit a project for cleanup opportunities
- Find only unused exports (smaller output)
- Check if a PR introduces quality risk
- Review a change and let fallow check the verdict
- Find code duplication
Key facts
- Installs on skills.sh
- 9,912 (Nov 2, 2026)
- Last repo commit
- Nov 1, 2026
- Audits on skills.sh
- Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass
What it includes
Finds unused files, exports, types, members and dependencies.
The core cleanup scan. Filter flags such as --unused-exports narrow the output.
Try: “Find unused exports in this repo.”
Runs code ·
fallow dead-codePass, warn or fail verdict for issues a change introduces, with --base <ref>.
The cheat sheet row for "commit or open a PR". The plugin commit gate runs it too.
Try: “Audit this PR for quality risk against main.”
Runs code ·
fallow auditReads a diff for orientation with --base <ref> --brief; always exits 0.
Runs code ·
fallow reviewComplexity hotspots, ownership, refactoring targets and coverage gaps.
Flags such as --hotspots --targets, --ownership and --coverage-gaps pick the report.
Runs code ·
fallow healthFinds duplicated code in strict, mild, weak or semantic mode.
Semantic mode detects renamed variables.
Try: “Find duplicated code and rank it.”
Runs code ·
fallow dupesDiscovers functions that may do the same thing despite different syntax.
SKILL.md says its output is discovery only, never proof of equivalent behavior.
Runs code ·
fallow similar-codeReports circular dependencies, boundary violations and policy violations.
Select one kind with --cycles, --boundaries or --policy.
Runs code ·
fallow architectureShows which architecture rules apply to a file before you change it.
Runs code ·
fallow guardTraces why an export or dependency is used, or how one module reaches another.
Paired with dead-code --trace before deleting an "unused" export or file.
Runs code ·
fallow traceFinds feature flag patterns.
Runs code ·
fallow flagsLists local security candidates for an agent to verify.
SKILL.md says it is not verified vulnerability scanning and does not prove exploitability.
Runs code ·
fallow securityExplains one issue type without running analysis.
Runs code ·
fallow explainRemoves supported unused exports and dependencies after --dry-run.
SKILL.md says to run --dry-run first, then fix --yes, which is required outside a terminal.
Try: “Preview then remove the unused exports.”
Runs code ·
fallow fixOpt-in local report of what fallow has surfaced over time.
impact enable is something the user owns. History is kept in the user config folder, not the repo.
Runs code ·
fallow impactReads production runtime data before an edit or delete.
Only when Fallow Cloud is connected, per SKILL.md.
Runs code ·
fallow coverage review-packetInitializes a new config file.
Runs code ·
fallow init
From the official docs
Permissions
Medium risk
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
fix can rewrite your code
SKILL.md tells the agent to run fix --dry-run first and then fix --yes to apply, and says --yes is required outside a terminal. Read the preview and work in a clean git tree. github.com
The plugin blocks commits and pushes
The plugin README says a PreToolUse hook on Bash runs fallow audit before git commit and git push and blocks the command on a fail verdict. It runs only in a project with a fallow config file or fallow dependency. Set FALLOW_PLUGIN_GATE=off to turn it off. github.com
The hook script reads your project files
fallow-gate.sh checks for .fallowrc.json, .fallowrc.jsonc, fallow.toml, .fallow.toml and package.json, and runs fallow from PATH, node_modules/.bin, yarn or npx --no-install. A missing fallow binary, missing jq or fallow below 2.85.0 lets the command through with a notice. github.com
The statusline command edits a Claude setting
The plugin README says it writes only after a preview and your confirmation, keeps the previous setting and will not overwrite later manual changes. The helper script also writes a copy of itself and a state file. github.com
Installing the CLI downloads code
The README says installing or running fallow through npm, npx or cargo may download the CLI. Telemetry is off unless you opt in with fallow telemetry enable. github.com
Technical details
| Turning on read-only | No read-only switch is documented. Analysis commands report findings; fix and init change files. |
|---|---|
| Talks to | npm registry or crates.io when the CLI is installed or run with npx. Fallow Cloud, only when you connect it and its commands are run |
| Sign-in and keys | No sign-in is needed for local analysis. Fallow Cloud commands need credentials that you configure. |
| Scopes |
|
Questions
What does the fallow skill do?
It teaches your assistant to run the fallow CLI on a TypeScript or JavaScript project. That finds unused code and dependencies, duplicates, circular imports, complexity hotspots and risky changes in a pull request.
How do I install it?
Use the install box at the top of this page. You also need the fallow tool itself in the project, because the skill only tells the assistant how to run it.
Does it need an API key?
Not for local analysis. SKILL.md and the README treat Fallow Cloud as an optional paid product, and its commands need credentials you configure.
Is it safe?
It runs a CLI in your repo and can delete unused exports and dependencies with fix --yes. The plugin route also adds a hook that can stop git commit and git push. Our advice: review the dry-run output and keep your tree clean in git.
What is the difference between the skills route and the plugin route?
The skills CLI route copies only the skill. The plugin route installs the fallow plugin, which the plugin README says holds 5 skills, a statusline command and the commit gate hook; Codex also gets the Fallow app.
Does it replace ESLint or tsc?
No. SKILL.md says not to use it for type checking, lint or format checks, verified security scanning or bundle size. It also has a section on migrating from knip or jscpd.
Is this the official fallow skill?
It is published in the fallow-rs/fallow-skills repo, which the README says holds agent skills for fallow and links to fallow-rs/fallow. The LICENSE names Bart Waardenburg, the same owner as in the plugin manifest.
Does it work in the Claude app?
The install box has no Claude app upload route, because the skill runs a CLI in your project folder.
Information
updated Nov 2, 2026
- Repository
- fallow-rs/fallow-skills
- Folder
fallow/skills/fallow
Sources
Docs, issue trackers and discussions this page is built from.
Show all 7Hide
- fallow-skills README · github.com
- fallow-skills marketplace.json · github.com
- fallow plugin.json · github.com
- fallow plugin README · github.com
- fallow hooks.json · github.com
- fallow-gate.sh · github.com
- fallow SKILL.md · github.com