---

---

The problem

Every agency repeats the same eight workflows every week.

Briefs get written from scratch. Approvals disappear into WhatsApp. Reports take two hours to format. Revision #3 gets done for free because nobody was counting. Quotes go out that quietly lose money because nobody checked the margin.

The solution

agencykit encodes 8 production-tested agency workflows as Claude Agent Skills. Install once, trigger with plain English (or Italian), and get complete, structured output ready to paste into any PM tool or send to a client.

Install

npx agencykit install

Or via curl:

curl -fsSL https://raw.githubusercontent.com/MarcoZorn/agencykit/main/install.sh | bash

The installer drops all 8 skills into .claude/skills/ (project-local if a .claude/ folder exists) or ~/.claude/skills/ (global). Then just describe what you need.

If agencykit saves you time, a ⭐ on GitHub helps others find it.

Skills

Skill	Trigger	What you get
project-brief	"new project", "nuovo progetto"	Full RECAP with task list, phases and risk register
client-approval	"approval", "approvazione"	Approval request with 48h SLA, escalation at 72h, revision counter
client-report	"monthly report", "report mensile"	A narrative client report — not a data dump
quote-builder	"quote", "preventivo"	Line-item quote + margin check (warns below 40%)
shooting-day	"shooting", "riprese"	ODG, shot list, equipment checklist, crew WhatsApp message
social-calendar	"content calendar", "piano editoriale"	4-week calendar + first week captions written in full
post-mortem	"post-mortem", "retrospective"	Structured retrospective appended to LESSONS.md
agency-status	"status update", "stato agenzia"	Traffic-light per project + morning briefing message

Full documentation: marcozorn.github.io/agencykit/skills

Brand voice

Every skill can speak in your agency's voice. Drop a BRAND.md file with your tone, terminology and examples, and each skill reads it before generating — so briefs, reports and client messages all sound like you, not like a default AI.

• Global voice — .claude/brand/BRAND.md applies to every skill.
• Per-skill voice — .claude/skills/[skill-name]/BRAND.md overrides the global one for that skill.
• No file? Skills fall back to a clean, professional default.

A BRAND.md typically holds your agency name and tone, a handful of example sentences in your voice, a list of words and phrases to avoid, and optional client-specific overrides. See BRAND.md.example for the full format.

# start from the template
mkdir -p .claude/brand
cp BRAND.md.example .claude/brand/BRAND.md

Example output

project-brief — Bloom Studio social media retainer

RECAP PROGETTO
Cliente: Bloom Studio
Servizio: Social media management (Instagram + TikTok) — 3-month retainer
Obiettivo: Grow Bloom Studio's Instagram from 8,200 to 12,000 engaged
           followers in 90 days and drive 40+ qualified DMs/month.
Deadline: 2026-10-15
Responsabile: Sara Kent
Priorità: High

FASI OPERATIVE
| Fase              | Responsabile | Output atteso                      | Deadline   |
| Discovery & audit | Sara Kent    | Channel audit, competitor teardown | 2026-07-18 |
| Content strategy  | Sara Kent    | 3-pillar strategy, KPI dashboard   | 2026-07-25 |
| Month 1 production| Luca Verdi   | 16 posts + 12 stories              | 2026-08-01 |

RISCHI E BLOCCHI
| Rischio                   | Soluzione                                       |
| Founder availability thin | Batch one shooting day up front                 |
| Slow approvals            | 48h SLA via client-approval, 72h escalation     |

Full version: examples/project-brief-output.md

quote-builder — 3-month package with margin check

CLIENT QUOTE — Bloom Studio · GBP
| Item                                | Qty | Unit price | Total  |
| Social strategy & channel audit     | 1   | £1,200     | £1,200 |
| Monthly content production (16/mo)  | 3   | £2,400     | £7,200 |
| Shooting day (batch production)     | 3   | £1,500     | £4,500 |
Total (inc. VAT): £18,720

INTERNAL VERSION (do not send)
⚠ MARGIN FLAG — Shooting days at 36% are below the 40% floor.
  To reach 40% margin: price each shooting day at £1,610 (+£110).
  Recommendation: raise the shooting day to £1,650 → overall margin 49% → ~52%.

Full version: examples/quote-builder-output.md

client-approval — Rossi Gioielli reel package (IT + EN)

See examples/client-approval-output.md — a full approval request in Italian and English, with the internal SLA/escalation note and the revision counter flagging the last included round.

Compatibility

Works with any agent that supports the agentskills.io spec:

Agent	Notes
Claude Code	Primary target — drop into .claude/skills/
Codex CLI	agentskills.io-compatible
Cursor	agentskills.io-compatible
Gemini CLI	agentskills.io-compatible
OpenClaw	agentskills.io-compatible

How it works

agencykit installs skill files into .claude/skills/ (project-level) or ~/.claude/skills/ (global). Each skill is a SKILL.md file with frontmatter, English instructions, and an Italian translation block. Your agent reads the skill and knows how to respond when you use the trigger phrase.

Usage

npx agencykit install            # install (local project, else global)
npx agencykit install --global   # force install to ~/.claude/skills
npx agencykit list               # list available skills
npx agencykit status             # show what's installed

Then, in Claude Code (or any compatible agent), just say:

"new project for Acme Corp, social media management, deadline end of month"

agencykit asks for the inputs it needs and returns a structured, paste-ready output.

Validate

Skills are checked in CI on every push. Run the validator locally:

node scripts/validate.js

Contributing

See CONTRIBUTING.md. PRs for new skills, bug fixes, or translations welcome.

License

MIT — use it, fork it, build on it. See LICENSE.

---

🇮🇹 Per le agenzie italiane

agencykit nasce dall'esperienza operativa reale nelle agenzie creative. Ogni skill codifica un flusso che il tuo team ripete ogni settimana: dal brief al report cliente, dalla scaletta dello shooting al piano editoriale, dal preventivo con controllo margine allo stato settimanale dell'agenzia.

Ogni skill è bilingue: inglese e italiano, con le frasi trigger in entrambe le lingue. Funziona con Claude Code, Codex CLI, Cursor, Gemini CLI e OpenClaw.

npx agencykit install

Le 8 skill vengono copiate nella tua cartella .claude/skills/. Da lì basta descrivere cosa ti serve — "nuovo progetto", "manda in approvazione", "report mensile", "preventivo", "scaletta shooting" — e ottieni un output strutturato, pronto da incollare in Asana, ClickUp o Notion.

---

Found this useful?

If agencykit saved you time writing a brief, chasing a client approval, or formatting a report — a ⭐ takes 2 seconds and helps other agencies find it.

→ Star on GitHub

---