Understanding Project Skills

Skills are loadable instruction sets that give agents specialized knowledge for complex, multi-step workflows. Unlike agents, skills are loaded on-demand—activated when needed, not always running.

Looking for Meta-Skills?

Meta-skills are a special category that generate project-specific skills based on your stack.

Learn about Meta-Skills

Skills vs Agents

While both provide AI capabilities, they serve different purposes:

Agents

•Autonomous entities that take actions
•Have their own system prompts and personalities
•Invoked via @agent or Task tool
•Run in separate sessions with fresh context

Skills

•Instruction sets that augment agent capabilities
•Provide step-by-step workflows for specific tasks
•Loaded into the current session when triggered
•Can include bundled scripts and templates

Think of it this way: An agent is like a specialist you hire. A skill is like a detailed recipe that specialist follows for a particular task.

When Skills Are Loaded

Skills are loaded dynamically based on context or explicit triggers:

Trigger Phrases

Each skill defines trigger phrases that activate it. When the agent detects these phrases in your request, it loads the relevant skill.

“create a prd”“fix merge conflicts”“take screenshot”

Context-Aware Loading

Some skills are loaded automatically based on what the agent encounters. When resolving merge conflicts, the merge-conflicts skill is automatically available.

Explicit Loading

Agents can explicitly load skills using the Skill tool when they need specialized instructions for a task.

skill(name: "prd")

Available Skills (62)

These skills provide specialized workflows for common development tasks:

adhoc-workflow

Ad-hoc mode workflow for Builder

ad-hoc modequick fix

agent-audit

Audit AI toolkit agents for compliance and coverage gaps

audit agentsagent compliance

agent-onboard

Onboard new or modified agents to ensure they are project-context aware and follow toolkit conventions

onboard agentnew agent setup

auth-config-check

Check authentication configuration before E2E tests

auth config checke2e auth

auth-generic

Generic/custom authentication implementation for any provider

custom authgeneric login

auth-headless

Headless authentication using API calls and session injection

headless authskip login ui

auth-nextauth-credentials

NextAuth credentials (email/password) authentication for AI agents

nextauth loginnextauth credentials

auth-supabase-otp

Supabase passwordless OTP authentication for AI agents

supabase otpsupabase magic link

auth-supabase-password

Supabase email/password authentication for AI agents

supabase passwordsupabase email login

browser-debugging

Browser debugging escalation protocol for when tests pass but the feature doesn't work in the user's browser

tests pass but doesn't workfeature broken in browser

builder-cli

CLI tool detection and proactive usage for Builder

cli commandproject cli

builder-dashboard

Dashboard templates for Builder session startup

builder-delegation

Sub-agent delegation patterns for Builder orchestration

delegate tocall subagent

builder-error-recovery

Error recovery patterns for Builder sessions

builder-verification

Verification handling for Builder sessions

critic-dispatch

Dispatch code reviews to appropriate critic agents based on file types

critic dispatchcode review routing

cve

Assess a CVE for exposure, exploitability, and remediation options across one or more repositories

deep-investigation

Hypothesis-driven, multi-track bug investigation

bug analysisroot cause investigation

dynamic-reassignment

Dynamic task reassignment on failure

agent failedreassign task

electron-build-deploy

Build, deploy, and relaunch Electron desktop apps before Playwright verification

test-flow Step 3.5 with apps.desktop.buildDeploy configured

git-sync

Team synchronization for multi-machine collaboration

team syncgit sync

helm-test-cleanup

Purge all Helm ADE test data for a repo from Supabase and local filesystem

clean up helmreset test data

human-voice

Review and fix AI-generated writing patterns

human voiceAI writing

marketing-copy

Generate marketing copy from product documentation and PRDs

write marketing copygenerate headlines

merge-conflicts

Resolve merge conflicts on a pull request by merging the target branch into the source branch and fixing conflicts

fix merge conflictsresolve conflicts

multi-session

Multi-session coordination for parallel AI sessions

oauth-callback-diagnostic

Diagnose OAuth callback URL mismatches for desktop and web apps authenticating to third-party services (GitHub, Google, Azure AD, etc

redirect_uri errorcallback URL mismatch

post-completion

Post-completion polish steps for Developer

post-completionpolish

prd

Generate a Product Requirements Document (PRD) for a new feature

create a prdwrite prd for

prd-to-json

Convert PRDs to prd

convert this prdturn this into developer format

prd-workflow

PRD mode workflow for Builder

PRD modebuild PRD

product-screenshots

Capture and maintain product screenshots for marketing and support use

capture screenshotsupdate product images

project-bootstrap

Bootstrap a new or existing project with stack detection and project

add projectnew project

project-scaffold

Generate project boilerplate from scaffold templates

scaffold projectgenerate boilerplate

public-page

Generate public-facing pages (marketing, legal, error, changelog)

create landing pagebuild pricing page

screenshot

Capture authenticated screenshots of web pages for visual verification

take screenshotcapture screenshot

self-correction

Self-correction loop for fixing test and build failures

fix loopself correction

session-handoff

Produce a Planner session handoff by writing the five structured fields to planner-state

session-log

Manage persistent session logs for Builder

session logsession state

session-monitor

Monitor an active Helm session by tailing the debug log file

monitor sessionwatch session

session-setup

Always-on session initialization for Builder

session-state

Session state management for resumability

session stateheartbeat

setup-auth

Interactive authentication configuration for AI agents

setup authconfigure authentication

spec-analyzer

Extract structured requirements from unstructured project specifications, PRDs, or feature descriptions into a RequirementsManifest

analyze specextract requirements

stack-advisor

Recommend technology stacks based on project requirements by scoring stack archetypes against a RequirementsManifest

recommend stackstack advice

start-dev-server

Start the development server for a project and wait for it to be ready

start dev serverstart server

test-doc-sync

Test documentation sync workflow for keeping test files and docs aligned

test doc syncsync test docs

test-failure-handling

Test failure logging, manual fallback, and skip flows

test failedskip test

test-flow

Unified test orchestrator — skip gate, activity resolution, quality check pipeline, completion prompt, and tier-2 skill loading

generate testsrun tests

test-prerequisite-detection

Detect and classify test prerequisites and blockers

prerequisite failureblocker detection

test-ui-verification

UI verification test patterns and flaky test handling

ui verificationflaky test

test-url-resolution

Resolve test base URL for E2E testing

test urlresolve test url

test-user-cleanup

Clean up dynamically created test users from the database

cleanup test usersdelete test accounts

test-verification-loop

3-pass stability verification and automated fix loop

stability checkverification loop

ui-test-electron

Playwright E2E testing patterns for Electron desktop applications

electron testdesktop app test

ui-test-flow

E2E test generation and execution workflow

run e2egenerate e2e

ui-test-full-app-audit

Comprehensive E2E test auditing workflow

e2e auditfull coverage audit

ui-test-ux-quality

Quality-beyond-correctness E2E testing patterns

flicker testvisual stability

ui-test-xcuitest

XCUITest patterns for native macOS and iOS apps

xcuitestui test swift

vectorize

Vectorize codebase and database schemas for semantic search

semantic searchvectorize codebase

vercel-supabase-alignment

Diagnose Vercel environment to Supabase project alignment issues

database errorenvironment mismatch

verification-contracts

Create verification contracts for task completion

verification contractsuccess criteria

Browse Project Skills Learn About Meta-Skills