AGENT SKILL

growth-architect-foundry

Multi-tenant agency growth engine with 31 deterministic commands. Use for ANY advertising, growth, campaign, content, lead, or client-reporting task. Triggers: foundry, audit, diagnose, report, campaign, content, tracking, leads, creative, competitive, season, forecast, tenants, calc, media buying, ad account, client report, growth engine, محرك النمو, تدقيق, تشخيص, تقرير, حملات, محتوى, تتبّع, نقطة التعادل, الروآس وقع, راجع حسابي, خطة, ميديا بلان, كبّر ولا أوقف

License 许可
MIT
最近更新
2026-11-14
Stars
★ 1.6k
Forks
491

01 前言 Front Matter

name
growth-architect-foundry
license
MIT
metadata
{ "repo": "https://github.com/mohamedOmar11111/growth-architect-foundry", "author": "Mohamed Omar", "version": "1.0.0" }
description
Multi-tenant agency growth engine with 31 deterministic commands. Use for ANY advertising, growth, campaign, content, lead, or client-reporting task. Triggers: foundry, audit, diagnose, report, campaign, content, tracking, leads, creative, competitive, season, forecast, tenants, calc, media buying, ad account, client report, growth engine, محرك النمو, تدقيق, تشخيص, تقرير, حملات, محتوى, تتبّع, نقطة التعادل, الروآس وقع, راجع حسابي, خطة, ميديا بلان, كبّر ولا أوقف

02 资源文件 4

03 SKILL.md 正文

Growth Architect Foundry

The multi-agent growth kernel for agencies who don't want to build one from scratch. 31 deterministic commands. Multi-tenant. Maker/checker safety. Arabic-first.

Step 0 — Intake (ask once, then work)

Before running any command, gather:

  1. Business model: COD e-com · prepaid e-com · lead-gen · app · B2B · local service
  2. Market(s) + language: country, dialect, expat segments
  3. Platforms + monthly budget (currency, VAT-inclusive?)
  4. Target: CPA/ROAS/CPL or unit economics to compute it. If missing → run calc breakeven first.
  5. Data: last 30 + previous 30 days by campaign (or "no data yet")

If enough numbers → answer first with assumptions, list missing at end. If basics missing → one-message intake.

Core Rules (every command)

  1. Signal before bids — broken tracking → fix first; other advice provisional
  2. Real economics — COD judged on delivered revenue ÷ spend
  3. Respect learning — ~50 conversion events/week; budget edits 15-20% at a time
  4. Consolidate before fragment — fewer campaigns with enough conversions beat many starved
  5. Creative = main targeting lever — diagnose angles, not fonts
  6. Waste first, scale second — search terms, placements, frequency, overlap
  7. Right comparison set — never judge Egyptian e-com by US lead-gen numbers
  8. Split by country + language — Egypt ≠ KSA ≠ UAE
  9. Every projection is an estimate with assumptions stated. Never guarantee
  10. Hands off live accounts — read free; any write needs explicit per-action yes

Output Contract

  • Lead with verdict (one line), then evidence
  • Numbers not adjectives; show calculation for ROAS/CPA/budget
  • Three action buckets: ⚡ do today · ? test this week · ? stop/reduce
  • Tables for plans/audits; naming: {country}_{platform}_{objective}_{audience}_{offer}_{yyyymm}
  • Reply in user's language/dialect; keep platform terms English (ROAS, CPA, CBO, Advantage+, PMax)
  • Mark each finding: confirmed by data or hypothesis — verify with X

Commands

Foundry home (the folder holding foundry.py): @@FOUNDRY_HOME@@ If that value starts with @@ the skill is being used in place: in Claude Code use ${CLAUDE_PLUGIN_ROOT}, in any other agent use the folder two levels above this SKILL.md.

All commands run via python "<home>/foundry.py" <group> <verb> [--client <tenant>] (python3 on macOS/Linux).

Workspace: run from the user's own project folder (or set FOUNDRY_WORKSPACE). Tenants, log.md, signals/ and tasks/ live THERE, never in the plugin folder. First time: setup doctor, then setup init --client <Name>.

Machine output: add --json to any command for one object: schema_version, command, ok, exit_code, data, output, report_path, stderr (data is structured for audit, diagnose, live pull and economics). Exit codes: 0 ok, 1 failed, 2 bad input, 3 refused by a safety gate. Parse data/ok, not the prose.

Failure rule: a FETCH FAILED line from live means the API call failed — never describe it as "no campaigns". Writes (pause, budget) are enforced in code: they need --confirm AND the user typing the campaign key at a real terminal, so an agent cannot complete them (exit code 3 = refused). Preview with the command without --confirm, show the user the exact change, and give them the full command to run themselves. Budget changes are capped at 20% cumulative per campaign per 7 days. Reports write to 04_Reports/ and append to log.md.

Account (deterministic, no LLM)

Command Arabic What it does
foundry account audit --client <t> راجع حسابي 0-100 score + grade + top fixes + 30-day plan
foundry account diagnose --client <t> الروآس وقع Single biggest constraint + ranked findings
foundry account report --client <t> --lang ar تقرير للعميل Decision-first client readout, EN or AR
foundry account report --client <t> --lang ar --format whatsapp — Same digest, WhatsApp-ready plain text

Economics (deterministic, zero-dep calculator)

Command Arabic What it does
foundry economics breakeven --price 499 --cogs 150 ... نقطة التعادل Break-even CPA + ROAS with stated assumptions
foundry economics cod --spend 20000 --orders 400 ... — Real ROAS on COD account + profit verdict
foundry economics learning --cpa 120 — Budget to exit learning phase (~50 events/week)
foundry economics vat --budget 10000 --country SA — VAT split by MENA country (rates last verified 2026-10)
foundry economics leadgen --deal-value 5000 --close-rate 0.25 ... — Break-even CPL / CPQL + funnel profit verdict for lead-gen funnels

Campaigns (reads captured ad-intel reports)

Command What it does
foundry campaigns list --client <t> All campaigns with status, spend, CTR, CPA
foundry campaigns status --client <t> Flagged: live vs paused vs zero-spend
foundry campaigns compare --client <t> Side-by-side with best CTR/CPA highlighted
foundry campaigns import --client <t> --file export.csv Import a platform CSV export into a standard ad-intel report (unblocks diagnose/campaigns/forecast with no API)

Content (reads signals, tasks, calendar)

Command What it does
foundry content pipeline --client <t> Full signal→brief→task→calendar pipeline view
foundry content calendar --client <t> Upcoming publish schedule

Tracking (reads .env, no API calls)

Command What it does
foundry tracking check --client <t> Credential readiness per integration
foundry tracking events --client <t> Optimization event audit per platform

Live (requires API credentials — touches real accounts)

Command Read/Write What it does
foundry live pull --client <t> Read Pull live campaign data from Google Ads + Meta
foundry live optimize --client <t> [--target-cpa <v>] Read Scale/hold/reduce/kill recommendations with numbers
foundry live pause --client <t> --campaign-id google:123 --confirm Write Pause a campaign (--confirm + human types key at a terminal)
foundry live budget --client <t> --campaign-id meta:456 --new-budget 500 --confirm Write Adjust budget (max 20% change, --confirm required)

Campaign IDs use platform:id format (e.g., google:12345678, meta:23456789). The live pull command outputs these keys.

Leads / Creative / Competitive / Seasonal / Forecast / Ops

Command What it does
foundry leads score --client <t> Lead task scoring breakdown
foundry creative hooks --client <t> Hook ideas from winning signals
foundry creative angles --client <t> Angle matrix from performance data
foundry competitive audit --client <t> Positioning from strategy directives
foundry season plan --client <t> MENA seasonal calendar + prep timelines
foundry forecast project --client <t> Budget projections from captured data
foundry ops status --client <t> Engine health overview
foundry ops cadence --client <t> Loop schedule compliance
foundry tenants list All configured tenants with readiness

Safety Rules

  1. Read-only default on live connections — MCP read free, any write needs explicit per-action yes with exact change shown
  2. Honest limits — thresholds are heuristics; verify changing facts; not legal/tax advice
  3. Confidence labels — every finding marked confirmed by data or hypothesis
  4. No auto-publish — maker/checker built into every task contract
  5. Money from calc.py — every money figure comes from deterministic code, not from the AI guessing

When to Use This Skill

Use this skill when the user mentions:

  • Advertising, ads, campaigns, media buying, performance
  • Account health, audit, diagnose, bottleneck, optimization
  • Client reports, weekly report, monthly report
  • Break-even, ROAS, CPA, COD, VAT, unit economics
  • Content pipeline, briefs, calendar, hooks, angles
  • Leads, scoring, outreach
  • Season, Ramadan, Eid, White Friday
  • Growth engine, foundry, محرك النمو
  • Any of the Arabic trigger phrases listed above

Installation

# Claude Code plugin
/plugin marketplace add mohamedOmar11111/growth-architect-foundry
/plugin install foundry@growth-architect-foundry

# OpenCode, Freebuff/Codebuff, Codex, Cursor (run from YOUR project folder)
python <path-to-foundry>/scripts/install_agent.py

# Standalone
python foundry.py account audit --client Aurora_Skincare

Honest Limits

  • Thresholds used in audit/diagnose are heuristics, not benchmarks — judge each account against its own 90-day baseline
  • VAT rates change — verify before relying on them
  • Seasonal dates shift ~10 days/year (lunar calendar) — verify before committing budget
  • Every projection is an estimate. Nothing here is a guarantee
  • This is not legal, tax, or financial advice

Content Vault

Ready-to-use content organized by type. All assets follow the templates/ASSET_TEMPLATE.md format.

Type Path Contents
Templates templates/ ASSET_TEMPLATE, CASE_STUDY_TEMPLATE, EXPERIMENT_LOG_TEMPLATE
Playbooks playbooks/ Launch (7-day onboarding), Funnel (ladder diagnosis), COD Operations
Frameworks frameworks/ COD Profit Funnel, Bilingual Content System
Benchmarks benchmarks/ MENA e-com benchmarks (sourced, cited)
Swipe Files swipe-files/ 20 hooks (EN+AR), CTAs, WhatsApp scripts
Worked Examples worked-examples/ COD ROAS diagnosis, Signal-to-content pipeline
Prompts prompts/ Account audit, Creative brief, Ad copy (MENA bilingual)

Every benchmark number is sourced. Every worked example is labeled fictional. Every prompt has an "honest limits" section.

04 同仓库更多 Skills

hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-review-with-annotations

Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-find-and-synthesize-context

Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-update-docs

Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-manage-records

Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-create-or-manage-interactive-html

Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.

★ 1.6k
hashgraph-online
hashgraph-online/awesome-codex-plugins

worktable-collaborate-threads

Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.

★ 1.6k