tuff

Capability lifecycle manager for coding agents.

A CLI for teams that keep agent capabilities version-controlled in the repo: install a capability, record where it came from, catch local edits before they cause a bad run, and emit the exact capability version each harness needs to behave as expected.

$uv tool install tuffcli

//how it works

The lifecycle Tuff manages

Agent capabilities are useful only if a team can tell what is installed, where it came from, and whether it changed.

Create or add

Start a new skill, tool, hook, or workflow, or bring existing agent files under management without rewriting their content.

Track

Record source, version, scope, harness, emitted path, and the clean install baseline in project state.

Diff

Compare the current files against the baseline so intentional edits are visible and accidental drift is not silent.

Update or check

Accept deliberate local changes, merge upstream updates where possible, and fail CI when tracked capabilities are not in the expected state.

~/tuff

$ tuff init

created tuff.lock

registered open-agents as the default harness

installed tuff-cli-guide

$ tuff create skill project-review

tracked .agents/skills/project-review/SKILL.md

$ tuff diff project-review

modified since baseline

$ tuff check

project-review skill open-agents ok

//capabilities

The things agents load should be first-class project assets

Tuff treats these as managed files, not magic settings hidden inside one developer's local harness.

SkillInstructions and references an agent can load for a project or task. ToolCallable scripts, functions, or MCP servers with a clear contract. HookAutomation attached to supported lifecycle events in a harness. WorkflowA repeatable sequence that composes capabilities for a larger job. Policy ControlGoverns all four. Set which capabilities can be installed, from which sources, and for which harnesses, enforced consistently across every agent your team runs.

//harnesses

Write once, emit per harness

Tuff compiles a canonical manifest into whatever file shape each agent harness expects natively.

type: hook

event: PreToolUse

runner: ./lint.sh

blocking: true

compile

claude-code .claude/hooks.json

cursor .cursor/hooks.yaml

codex .codex/hooks.json

//day two

Operational Features

The parts that matter after day one.

01CI Integration

Run the check in CI and fail the build on drift, not after it's already shipped.

GitHub Actions

02Tuff Skills

Tuff ships as a skill itself. The agent learns the CLI and hook spec on install, so it writes manifests and runs commands for you instead of you doing it by hand.

tuff init

03Audit Reports

Generate a full report of every installed capability, its source, and its policy status, right from the CLI, no hosted service required.

tuff generate report