All skills
Skillintermediate

CLAUDE.md

You're the orchestrator. Spawn subagents via `Task()` for focused work.

Claude Code Knowledge Pack7/10/2026

Overview

CLAUDE.md

You're the orchestrator. Spawn subagents via Task() for focused work.

Plugin mode: When installed as a plugin, agents and skills are available natively. Use /full-audit, /pre-commit, /pre-deploy, /new-feature, /bug-fix, /release-prep skills.

Agents (24 Total)

Audit Agents (11) - NON-OVERLAPPING SCOPE

TypePromptDoes
bug-auditoragents/bug-auditor.mdRuntime bugs (NOT security)
code-auditoragents/code-auditor.mdCode quality (NOT security/bugs)
security-auditoragents/security-auditor.mdALL security (single authority)
doc-auditoragents/doc-auditor.mdDocumentation
infra-auditoragents/infra-auditor.mdConfig/infra
ui-auditoragents/ui-auditor.mdUI/UX & a11y
db-auditoragents/db-auditor.mdDatabase & queries
perf-auditoragents/perf-auditor.mdPerformance
dep-auditoragents/dep-auditor.mdDependencies
seo-auditoragents/seo-auditor.mdSEO & meta tags
api-testeragents/api-tester.mdAPI validation

Fix/Implement Agents (4)

TypePromptDoes
fix-planneragents/fix-planner.mdPrioritize fixes
code-fixeragents/code-fixer.mdImplement fixes
test-runneragents/test-runner.mdRun tests
test-writeragents/test-writer.mdGenerate tests

Browser QA Agents (4)

TypePromptDoes
browser-qa-agentagents/browser-qa-agent.mdChrome UI testing
fullstack-qa-orchestratoragents/fullstack-qa-orchestrator.mdFind-fix-verify loop
console-monitoragents/console-monitor.mdReal-time console
visual-diffagents/visual-diff.mdScreenshot comparison

Deploy Agents (2)

TypePromptDoes
deploy-checkeragents/deploy-checker.mdPre-deploy validation
env-validatoragents/env-validator.mdEnvironment config

Utility Agents (2)

TypePromptDoes
pr-writeragents/pr-writer.mdPR descriptions
seed-generatoragents/seed-generator.mdTest data

Supervisors (1)

TypePromptDoes
architect-revieweragents/architect-reviewer.mdFinal approval

Spawning

Single:

Task(subagent_type="bug-auditor", prompt="...", description="Security")

Parallel (same response):

Task(subagent_type="bug-auditor", prompt="...", description="Security")
Task(subagent_type="code-auditor", prompt="...", description="Code")
Task(subagent_type="security-auditor", prompt="...", description="OWASP")
Task(subagent_type="db-auditor", prompt="...", description="Database")

Sequential: Wait for result, then next Task.

Workflows / Skills

Available as slash commands when installed as a plugin, or see workflows/ for details.

WorkflowPurposeAgents
pre-commitBefore committingcode-auditor, test-runner
pre-deployBefore deploymentdeploy-checker, env-validator, dep-auditor
full-auditComplete auditAll 11 auditors → fix-planner
new-featureBuild featuretest-writer, code-fixer, test-runner, browser-qa
bug-fixFix bugstest-writer, code-fixer, test-runner
release-prepRelease prepfull-audit, fixes, deploy-checker, pr-writer

Quick commands:

  • Full audit: Spawn all auditors parallel → fix-planner
  • Fix cycle: fix-planner → code-fixer → test-runner → architect-reviewer
  • Quick check: Just bug-auditor
  • Browser QA: browser-qa-agent → fix-planner → code-fixer → verify

Outputs

All go to .claude/audits/:

FileSource
AUDIT_SECURITY.mdsecurity-auditor (SINGLE security authority)
AUDIT_BUGS.mdbug-auditor (runtime bugs only)
AUDIT_CODE.mdcode-auditor (quality only)
AUDIT_DOCS.mddoc-auditor
AUDIT_INFRA.mdinfra-auditor
AUDIT_UI_UX.mdui-auditor
AUDIT_DB.mddb-auditor
AUDIT_PERF.mdperf-auditor
AUDIT_DEPS.mddep-auditor
AUDIT_SEO.mdseo-auditor
API_TEST_REPORT.mdapi-tester
DEPLOY_CHECK.mddeploy-checker
ENV_REPORT.mdenv-validator
FIXES.mdfix-planner (with deduplication)
TEST_REPORT.mdtest-runner
AUDIT_BROWSER_QA.mdbrowser-qa-agent
QA_SESSION_LOG.mdfullstack-qa-orchestrator
QA_COMPLETE.mdfullstack-qa-orchestrator
EXECUTION_LOG.mdAll agents (status tracking)

Agent Status Protocol

Every agent output MUST start with a status block:

---
agent: [agent-name]
status: COMPLETE | PARTIAL | SKIPPED | ERROR
timestamp: [ISO timestamp]
duration: [seconds]
findings: [count]
errors: []
skipped_checks: []
---

Project

Stack: Critical paths: Don't touch: .env