_ _ _ _____ _
/_\ __ _ ___ _ _ __| |_(_)__ |_ _|__ ___| |___
/ _ \/ _` / -_) ' \/ _| _| / _| | |/ _ \/ _ \ (_-<
/_/ \_\__, \___|_||_\__|\__|_\__| |_|\___/\___/_/__/
|___/
Portable Agent Skills, workflows, and automation for AI-assisted development
Independent Community Project — No official affiliation with Anthropic, OpenAI, Oracle, or any other mentioned product team. Skills are community-created and use the portable SKILL.md Agent Skills format.
What This Is
agentic-tools is a curated library of 41 Agent Skills plus Claude Code agents, hooks, and workflow templates. It is built around the late-May 2026 Agent Skills model: each skill is a focused SKILL.md package with trigger-oriented metadata, progressive disclosure, optional scripts or references, and validation gates that keep install surfaces in sync.
The library has two jobs:
- Provide a practical development pipeline from requirements to review.
- Package repeatable domain expertise, including a distinct OCI and Oracle skill pack, in a portable format that can be reused by skills-compatible agents.
The core development pipeline:
/prd → /prd --to-plan → /orchestrate → agents use /implement → /review-all → /health-check → PR
Agent Skills Standard
As of late May 2026, Agent Skills are a lightweight open format built around a directory containing a required SKILL.md file plus optional scripts/, references/, and assets/. Agents use progressive disclosure: they first see only the name and description, load the full SKILL.md when the task matches, and load bundled resources only when needed.
This repo follows that model with a few local quality gates:
- Every registered skill package under
skills/ is installable and must contain SKILL.md.
name must match the package folder name; nested OCI skills use IDs like oci/zpr-security.- Descriptions are trigger-oriented and include quoted user phrases for reliable discovery.
- Skill bodies include load boundaries, anti-patterns, and an
## Arguments section.
- Large or high-drift material belongs in focused
references/ files, not in the always-loaded body.
scripts/ci/validate-skill-library.mjs checks skills, registries, READMEs, referenced scripts, and the OCI manifest together.
References: Agent Skills overview, Agent Skills specification, Agent Skills best practices, Claude custom Skills guide, and OpenAI Academy: Skills.
Packages
| Package | What it includes |
|---|
| agentic-tools (this repo) | Everything — all skills, agents, hooks, configs |
| team-pipeline | Pipeline skills + agents only — for agent team orchestration |
# Install everything
npx agentic-tools init
# Install just the pipeline
npx @acedergren/team-pipeline init
Quick Start
Via npx (recommended)
# Install into current project
npx agentic-tools init
# Install into a specific project
npx agentic-tools init ./my-project
# See what's included
npx agentic-tools list
Via git clone
git clone https://github.com/acedergren/agentic-tools.git
cd agentic-tools
./install.sh /path/to/your/project
This copies skills to .claude/skills/, agents to .claude/agents/, and hook examples to .claude/hooks/.
Install Individual Skills
# Copy a single skill from a cloned repo
cp -r skills/implement /path/to/project/.claude/skills/
# Or symlink for development
ln -s $(pwd)/skills/implement /path/to/project/.claude/skills/
Validate the Library
# Validate all skills, installer coverage, docs coverage, scripts, and OCI manifest
npm run skills:validate
# Run the full local skill CI: validation, install smoke, and secret scan
npm run skills:ci
Try It
# In your project with Claude Code:
/implement add validation to the user signup endpoint
/tdd add rate limiting to POST /api/search
/bugfix "TypeError: Cannot read property 'userId' of undefined"
/migrate "@old/pkg/utils → @new/pkg/core"
/health-check
/review-all
/prod-readiness
/oci route this OCI architecture review to the right specialist skills
Skills
Development Pipeline
These skills chain together into a full implementation workflow.
| Skill | What It Does | Key Feature |
|---|
| /implement | End-to-end feature pipeline | Pre-flight → TDD → scope guard → commit |
| /tdd | Test-driven development cycle | Mock bootstrap phase catches wiring issues early |
| /write-tests | Add tests to existing code | Module-type-aware mock strategy selection |
| /quality-commit | Quality gates + commit | Lint, typecheck, test, then commit |
| /bugfix | Autonomous bug fix pipeline | Evidence → hypothesis → minimal fix → verify → commit, no hand-holding |
| /migrate | Codebase migration orchestrator | Dedup check → scripted bulk sed → typecheck → residual scan → commit |
Review & Quality
| Skill | What It Does | Key Feature |
|---|
| /review-all | Parallel multi-reviewer pipeline | Security + API audit + scope check agents |
| /health-check | Full codebase diagnostic | 7+ gates: types, tests, lint, security, dead code |
| /api-audit | Route-to-type contract validation | Finds auth gaps, missing schemas, type drift |
| /doc-sync | Documentation drift detection | Compares docs against actual codebase state |
| /prod-readiness | 5-agent production readiness review | Security + test coverage + perf + observability + code quality → prioritized report |
Planning & Orchestration
| Skill | What It Does | Key Feature |
|---|
| /prd | PRD creation with validation gates | 10-gate validation checklist, --to-plan mode |
| /orchestrate | Multi-agent team coordination | Wave-based execution, heartbeat monitoring |
| /phase-kickoff | Phase scaffolding | Branch + test shells + roadmap entry |
Content & Utilities
| Skill | What It Does | Key Feature |
|---|
| /shadcn-svelte | shadcn-svelte architecture | Tailwind v4.1 patterns, library selection |
| /refactor-module | Terraform module extraction | Decision framework for when to extract |
| /write-natural-swedish | Natural Swedish writing | Contemporary Swedish + tech-company presets |
OCI and Oracle Skill Pack
The OCI skills stay individually installable while skills/oci provides the visible ownership boundary for Oracle-related work. All OCI and Oracle-related specialists live under skills/oci/<skill-name>/ and use skill IDs like oci/zpr-security. The manifest is validated in CI against skill metadata, the CLI, and the Bash installer.
See the OCI Skill Catalog and Use Cases for concrete example prompts, routing guidance, and multi-skill workflows.
| Skill | What It Does | Key Feature |
|---|
| /oci | OCI and Oracle skill-pack router | Canonical manifest for separation of duties |
| /oci/best-practices | OCI architecture review router | Cross-domain triage into specialist skills |
| /oci/compute-management | OCI Compute operations | Shapes, capacity, instance principals, boot volumes, and access posture |
| /oci/networking-management | OCI Networking operations | VCN, DRG, FastConnect, VPN, NSG, security lists, and private connectivity |
| /oci/iam-identity-management | OCI IAM and identity | IAM policies, identity domains, IDCS compatibility, dynamic groups, and 403s |
| /oci/infrastructure-as-code | OCI Terraform hub | Native OCI backend, auth, import, drift, modules, and realm guardrails |
| /oci/oci-resource-manager | OCI Resource Manager operations | Stacks, jobs, state, source providers, private endpoints, and IAM boundaries |
| /oci/oci-security-control-plane |
Complete Skill Directory
The install surfaces expose every registered skill package under skills/, including nested OCI skill IDs. CI validates this table, bin/cli.js, install.sh, and each skill directory together.
| Skill | Trigger Summary |
|---|
| /api-audit | Use when auditing API routes for schema drift, missing auth, or validation gaps. Scans routes against shared TypeScript |
| /oci/best-practices | "review OCI architecture", "avoid OCI anti-patterns", "plan an Oracle Cloud migration", "evaluate OCI Well-Architected risks", or "choose which OCI skill applies" |
| /bugfix | Use when given a bug report, failing test, stack trace, or CI failure that needs diagnosis and repair. Autonomously diag |
| /oci/compute-management | "launch OCI compute", "choose an OCI shape", "debug compute capacity", "configure instance principals", or "optimize OCI instance cost" |
| /oci/database-management | "create an OCI database", "choose DB System vs ADB", "manage PDB lifecycle", "route ADB work", or "plan Oracle database provisioning" |
| /doc-sync | Use when auditing or fixing drift between project documentation and the actual codebase. Detects stale architecture diag |
| /oci/fastify-better-auth-bridge | "bridge Better Auth into Fastify", "fix Fastify session resolution", "forward auth cookies", "patch IDCS org context", or "decorate Fastify request auth" |
| /oci/finops-cost-optimization |
Agents
Custom agents for specialized tasks. Use with Claude Code's Task tool.
| Agent | Model | Purpose |
|---|
| mock-debugger | Sonnet | Diagnoses Vitest mock wiring failures — 6 anti-pattern checklist |
| security-reviewer | Opus | OWASP Top 10 review with file:line references |
Using Agents
Agents are automatically available when placed in .claude/agents/. Claude Code uses them via the Task tool:
# In Claude Code, the /review-all skill automatically spawns the security-reviewer agent
# Or reference directly: "Use the mock-debugger agent to diagnose why my tests fail"
Hooks
Pre-built automation that runs on Claude Code tool events. See claude/hooks/README.md for full documentation.
Examples (5 production-ready)
| Hook | Trigger | What It Does |
|---|
commit-preflight.sh | Pre-Bash (git commit) | Lint + typecheck staged files |
security-scan.sh | Pre-Bash (git push) | Semgrep security scan |
cost-tracker.sh | Post-Bash | Track cloud resource costs |
session-summary.sh | Session end | Log session activity |
file-backup.sh | Pre-Edit | Backup files before editing |
Templates (4 starters)
Quick-start templates for building your own hooks: bash-pre.sh, edit-pre.sh, session-start.sh, user-prompt-submit.sh.
Configuration
User-Level Permissions (claude/settings/)
Global permissions covering 400+ commands across 30+ categories:
- Languages: Node.js, Python, Go, Rust, Java, Ruby, PHP, C/C++, Swift
- Cloud: AWS, GCP, Azure, OCI, Cloudflare
- DevOps: Docker, Kubernetes, Terraform, Ansible
- Testing: Vitest, Jest, Playwright, Cypress
- Security: Semgrep, Snyk, Trivy
# Install global permissions
cp claude/settings/user-level/config.json ~/.config/claude/config.json
Git Workflows (git-workflow/)
- 3 workflow strategies (Trunk-Based, GitHub Flow, GitFlow)
- 8 commit templates (Conventional Commits)
- 6 PR templates
- Branch naming conventions
The Pipeline
How the skills work together for a complete feature lifecycle:
┌─────────────────────────────────────────────────────────────┐
│ PLANNING PHASE │
│ │
│ /prd ──────────► PRD.md ──────────► /prd --to-plan │
│ (interactive (validated (generates │
│ discovery) requirements) task plan) │
└────────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ EXECUTION PHASE │
│ │
│ /orchestrate ──► spawn agent team ──► each agent runs: │
│ (coordinates (haiku/sonnet /implement │
│ waves) per task) (full TDD pipeline) │
│ │
│ /implement phases: │
│ 0. Pre-flight (git clean? types compile? tests green?) │
│ 1. Understand & Plan │
│ 2. Bootstrap Mock (1 test to validate wiring) │
│ 3. Red — write failing tests │
│ 4. Green — minimum implementation │
│ 5. Scope Guard (revert out-of-scope changes) │
│ 6. Full Suite + Quality Gates │
│ 7. Commit │
└────────────────────────┬────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ REVIEW PHASE │
│ │
│ /review-all ──► parallel agents: │
│ • security-reviewer (OWASP Top 10) │
│ • API route auditor (schema coverage) │
│ • scope auditor (drift detection) │
│ │
│ /health-check ► types ✓ tests ✓ lint ✓ security ✓ │
│ dead code ✓ circular deps ✓ audit ✓ │
│ │
│ Verdict: READY TO MERGE / NEEDS FIXES / NEEDS DISCUSSION │
└─────────────────────────────────────────────────────────────┘
Customization
These skills are designed to be adapted to your project. Common customizations:
Project-Specific Paths
Update file paths in skills to match your project structure:
# In skills/health-check/SKILL.md, update workspace paths:
cd apps/api && npx tsc --noEmit → cd src && npx tsc --noEmit
Additional Review Agents
Add project-specific reviewers to /review-all:
# In skills/review-all/SKILL.md, add to the agent table:
| **DB Reviewer** | `oracle-query-reviewer` (custom) | Changed SQL files | Bind params, atomic ops |
Custom Quality Gates
Extend /health-check with project-specific checks:
# Add to skills/health-check/SKILL.md:
### 8. OpenAPI Lint
npx spectral lint openapi.json --ruleset .spectral.yaml
Repository Structure
agentic-tools/
├── README.md
├── package.json # npm package + skill CI scripts
├── install.sh # Shell installer
├── LICENSE
│
├── bin/
│ └── cli.js # npx agentic-tools init/list/help
│
├── scripts/ci/ # Skill library validation gates
│ ├── validate-skill-library.mjs
│ ├── install-smoke.mjs
│ ├── secret-scan.mjs
│ └── run-all.mjs
│
├── skills/ # 41 Agent Skills, registered from top-level and OCI nested packages
│ ├── README.md # Skill catalog and standards
│ ├── oci/ # OCI/Oracle ownership boundary
│ │ ├── SKILL.md # OCI skill-pack router
│ │ ├── README.md
│ │ ├── manifest.json # Canonical OCI skill inventory
│ │ ├── best-practices/ # OCI architecture router
│ │ ├── infrastructure-as-code/ # OCI Terraform, state, import, and security automation
│ │ ├── oracle-dba/ # Autonomous AI Database operations
│ │ ├── oci-resource-manager/ # OCI Resource Manager operations
│ │ ├── oci-security-control-plane/
│ │ ├── managed-bastion-access/
│ │ ├── zpr-security/
│ │ └── ... # Remaining OCI and Oracle-related packages
│ ├── implement/ # Full TDD feature pipeline
│ ├── review-all/ # Parallel review pipeline
│ ├── health-check/ # Codebase diagnostics
│ └── ... # Remaining skill packages
│
├── claude/
│ ├── agents/ # Custom agent definitions
│ │ ├── mock-debugger.md
│ │ └── security-reviewer.md
│ ├── hooks/ # Hook scripts + templates
│ │ ├── README.md
│ │ ├── examples/ # 5 production-ready hooks
│ │ └── hook-templates/ # 4 starter templates
│ └── settings/ # Permission configs
│ ├── user-level/config.json # 400+ command permissions
│ └── templates/ # Platform-specific configs
│
└── git-workflow/ # Git strategy guides
├── workflows/ # 3 workflow strategies
├── commit-templates/ # 8 conventional commit templates
├── pr-templates/ # 6 PR templates
└── branch-naming/ # Naming conventions
Contributing
Found a missing pattern or have a battle-tested skill? PRs welcome.
- Put general skills at
skills/<skill-name>/SKILL.md; put all OCI or Oracle-related skills under skills/oci/<skill-name>/SKILL.md. Keep the package folder name and frontmatter name identical.
- Write a trigger-oriented
description with concrete quoted user phrases.
- Include
## When to Use, ## Do NOT load this skill when, anti-patterns, and ## Arguments.
- Keep the always-loaded body lean; move detailed docs, volatile facts, examples, and API notes into
references/.
- Add scripts only when they provide deterministic value, and make sure every referenced script exists.
- Update
bin/cli.js, install.sh, README.md, and skills/README.md; every Oracle-related skill also belongs in skills/oci/manifest.json.
- Run
npm run skills:ci before opening a PR.
See CONTRIBUTING.md for full guidelines.
License
MIT License — see LICENSE for details.
Legal Disclaimer: This is an independent community project with no official affiliation, partnership, or endorsement from Anthropic or any other mentioned products or services. References indicate compatibility, not affiliation.
Last Reviewed: May 27, 2026
Compatibility: Agent Skills SKILL.md format; local installers target Claude Code .claude/skills/ by default.
