Agentic Workflows

Repo-native operating files for controlled AI, operator, and growth marketing work.

Prompts are not the product. The product is the operating loop: context, delegation, verification, approval, artifact, learning.

What this is

agentic-workflows turns AI workflows into repo-native operating files: validate them, render runbooks, audit authority, and compile them into agent skills. The current sample pack focuses on public-safe operator workflows and growth marketing launch work: workspace operations, paid media, analytics consent, SEO, product marketing context, and social content review.

It is still a playbook, but v2 adds a runnable foundation:

• .workflow.yml files for executable-style workflow definitions
• a machine-readable workflow schema
• a Bun CLI for validation, runbook rendering, authority audit, and scaffolding
• repo templates for AGENTS.md and skill drafts

The emphasis is not on clever phrasing. The emphasis is on the operating system around the agent:

• what context the agent receives
• what authority it has
• how work is delegated
• what must be verified
• which actions require human approval
• what artifact is produced
• what lesson is saved for next time

This is not a prompt dump. Each workflow includes:

• when to use it
• required context
• authority boundaries
• expected artifact
• verification gate
• failure modes
• public-safe synthetic examples

Quick start

Requirements:

• Bun

Run the CLI from the repo root:

bun run validate
bun cli/aw.ts inventory
bun cli/aw.ts check
bun cli/aw.ts check-skills
bun cli/aw.ts catalog-check
bun cli/aw.ts eval-check
bun cli/aw.ts publication-scan
bun cli/aw.ts publication-scan --list
bun cli/aw.ts runbook workflows/repo-triage.workflow.yml
bun cli/aw.ts audit workflows/research-to-decision.workflow.yml
bun cli/aw.ts new workflow customer-feedback-triage
bun cli/aw.ts new skill launch-readiness

CLI commands:

aw validate <workflow>
aw check [workflow...]
aw check-skills [skill...]
aw catalog-check
aw eval-check [fixture...]
aw publication-scan [--list] [file...]
aw inventory
aw runbook <workflow>
aw audit <workflow>
aw new workflow <name>
aw new skill <name>

Five-minute tour

If you are evaluating the repo, run these in order:

bun run validate
bun cli/aw.ts inventory
bun cli/aw.ts runbook workflows/repo-triage.workflow.yml
bun cli/aw.ts audit workflows/external-action-gate.workflow.yml

That gives you:

1. a full validation gate for tests, workflows, skills, and publication safety
2. an inventory of the workflow, skill, and example surface area
3. a rendered runbook for repo triage
4. an authority audit for external-write approval boundaries

Start with Fictional product audit for the best end-to-end example of the operating loop.

Who this is for

• founders, operators, chiefs of staff, and product leads adopting AI agents
• growth marketers, paid-media operators, and lifecycle teams adding AI to launch, tracking, and approval workflows
• engineering and product teams designing human-in-the-loop workflows
• people evaluating what competent AI delegation looks like in practice
• anyone who wants less AI theater and more reliable AI-assisted execution

Growth marketer quick path

For a launch or operator review, start with the smallest workflow that matches the real risk:

1. Build stable product and audience context with product-marketing-context-builder.
2. Check tracking and consent with analytics-consent-audit.
3. Review ad claims and launch state with ad-preflight-review and paid-social-launch-gate.
4. Use google-workspace-operator-pack when the agent needs a draft-first SMB operating layer across Sheets, Drive/Docs, Calendar, and Gmail.
5. Use meta-ads-cli-dry-run-adapter when planning Meta Ads CLI or Marketing API work without account access by default.

The examples under examples are intentionally fictional. They show the artifact shape and approval boundaries without publishing real accounts, IDs, budgets, credentials, customer data, or private workspace state.

The operating loop

flowchart LR
  A[Context] --> B[Delegation]
  B --> C[Execution]
  C --> D[Verification]
  D --> E{Approval needed?}
  E -- No --> F[Artifact]
  E -- Yes --> G[Human approval]
  G --> F
  F --> H[Learning]
  H --> A

Loading

A good agentic workflow is a loop, not a one-off prompt. It starts with scoped context, delegates work with clear authority, verifies the result, gates risky actions, produces an artifact, and captures reusable lessons.

Core principles

1. Artifacts over prompts — a workflow should end in something durable: a memo, PR, issue tree, checklist, test result, decision log, or report.
2. Human-controlled authority — AI may draft, inspect, summarize, and propose; humans approve external writes, destructive actions, and sensitive decisions.
3. Main-agent accountability — parallel agents can help, but the operator synthesizes, verifies, and owns the recommendation.
4. Private by default — workflows are templates. Examples are synthetic. Private memory, secrets, client work, and internal context do not belong here.
5. Learning compounds — save reusable procedures and failure modes, not raw transcripts.

Executable workflow files

Executable-style examples live beside the markdown playbooks:

Workflow file	Use it for	Try it
repo-triage.workflow.yml	Mapping an unfamiliar repo before edits	bun cli/aw.ts runbook workflows/repo-triage.workflow.yml
research-to-decision.workflow.yml	Research that must end in a recommendation	bun cli/aw.ts audit workflows/research-to-decision.workflow.yml
external-action-gate.workflow.yml	Preparing an external write for approval	bun cli/aw.ts runbook workflows/external-action-gate.workflow.yml
google-workspace-operator-pack.workflow.yml	Mapping a draft-first SMB operator across Sheets, Drive/Docs, Calendar, and Gmail	bun cli/aw.ts runbook workflows/google-workspace-operator-pack.workflow.yml
analytics-consent-audit.workflow.yml	Auditing consent-gated analytics and conversion tracking	bun cli/aw.ts runbook workflows/analytics-consent-audit.workflow.yml
google-ads-upload-qa.workflow.yml	Reviewing Google Ads bulk uploads before posting account changes	bun cli/aw.ts audit workflows/google-ads-upload-qa.workflow.yml
ad-preflight-review.workflow.yml	Reviewing ad copy, claims, landing-page alignment, and launch approvals	bun cli/aw.ts audit workflows/ad-preflight-review.workflow.yml
meta-ads-cli-dry-run-adapter.workflow.yml	Preparing Meta Ads CLI or Marketing API work without default account access	bun cli/aw.ts runbook workflows/meta-ads-cli-dry-run-adapter.workflow.yml
paid-social-launch-gate.workflow.yml	Gating paid-social submission, enablement, event changes, and spend scaling	bun cli/aw.ts audit workflows/paid-social-launch-gate.workflow.yml
technical-seo-launch-audit.workflow.yml	Auditing crawl, indexation, metadata, sitemap, robots, and schema launch readiness	bun cli/aw.ts runbook workflows/technical-seo-launch-audit.workflow.yml
product-marketing-context-builder.workflow.yml	Building stable product, audience, proof, and claim-boundary context	bun cli/aw.ts runbook workflows/product-marketing-context-builder.workflow.yml
growth-loop-diagnosis.workflow.yml	Diagnosing growth loops, weak links, confidence, and next experiments	bun cli/aw.ts audit workflows/growth-loop-diagnosis.workflow.yml
social-content-fact-check-rewrite.workflow.yml	Reviewing and rewriting social posts before publication	bun cli/aw.ts audit workflows/social-content-fact-check-rewrite.workflow.yml
growth-launch-readiness.workflow.yml	Chaining context, creative, launch, analytics, approval, and learning checks	bun cli/aw.ts runbook workflows/growth-launch-readiness.workflow.yml

The schema is schema/workflow.schema.json.

Each workflow declares:

• name
• goal
• trigger
• inputs
• allowed_tools
• authority
• risk_level
• required_permissions
• external_side_effects
• destructive_actions
• dry_run
• approval_required
• steps
• verification
• artifacts
• memory_update

Operator and growth skills

The skills/ directory contains repo-native skill drafts for repeatable operator and growth marketing workflows. These skills are public-safe operating files, not private prompt dumps. Each skill names its inputs, authority boundary, approval gates, verification gate, and output artifact. check-skills also rejects a small set of obvious publication-policy violations such as private home paths and common token shapes.

Current skills:

Skill	Use it for
google-workspace-operator-pack	Designing a draft-first SMB operator layer across Sheets, Drive/Docs, Calendar, and Gmail
analytics-consent-audit	Auditing consent state, tag loading, conversion-event dispatch, and attribution evidence
google-ads-upload-qa	Reviewing Google Ads bulk upload packages before posting account changes
ad-preflight-review	Reviewing ad copy, claims, landing-page alignment, and approval requirements before launch
meta-ads-cli-dry-run-adapter	Designing dry-run Meta Ads CLI and Marketing API command plans with account-access gates
paid-social-launch-gate	Verifying paid-social launch readiness before submission, enablement, or scaling
technical-seo-launch-audit	Checking crawl, indexation, metadata, sitemap, robots, and schema launch readiness
product-marketing-context-builder	Building stable product, audience, proof, and claim-boundary context for growth work
growth-loop-diagnosis	Diagnosing the current growth loop, weakest link, confidence, and next experiment
social-content-fact-check-rewrite	Fact-checking and rewriting social posts before publication

Use growth skill eval prompts, operator skill eval prompts, and the machine-readable eval fixtures beside them to test skills with synthetic Acme Sleep or Acme Repair scenarios and a shared scoring rubric.

Validate skills and public-facing files with:

bun cli/aw.ts check-skills
bun cli/aw.ts catalog-check
bun cli/aw.ts eval-check
bun cli/aw.ts publication-scan
bun cli/aw.ts publication-scan --list

Authority levels

Level	Meaning
read_only	Inspect, summarize, and recommend. Do not modify files or external systems.
local_write	Modify files in the local repo or workspace. Do not write to external systems.
external_draft	Draft external-facing messages, issues, posts, or PR text. Do not send or publish.
external_write_requires_approval	Prepare an external write, then stop for explicit human approval before execution.
destructive_forbidden	Destructive actions are outside scope, even with tool access.

Markdown playbooks

Workflow	Use it for	Output artifact
Agentic project pulse	Project health and next actions	Pulse memo
Approval boundary matrix	Defining what agents may do	Authority table
Subagent delegation brief	Parallel task delegation	Brief + result spec
Multi-agent review loop	Research/review/design sprints	Synthesized recommendation
External action gate	Sending/posting/commenting/publishing	Approval checklist
Google Workspace operator pack	Mapping a draft-first SMB operating layer across Sheets, Drive/Docs, Calendar, and Gmail	Operator map
Analytics consent audit	Checking consent-gated analytics and conversion tracking	Audit report
Google Ads upload QA	Reviewing paid-search bulk uploads before account changes	QA report
Ad preflight review	Reviewing ad claims, landing-page alignment, and launch approvals	Preflight report
Meta Ads CLI dry-run adapter	Preparing Meta Ads CLI or Marketing API work without default account access	Adapter map
Paid social launch gate	Checking paid-social launch readiness before platform-visible changes	Launch gate report
Technical SEO launch audit	Checking crawl, indexation, sitemap, robots, metadata, and schema readiness	SEO audit report
Product marketing context builder	Building reusable product, audience, proof, and claim-boundary context	Context document
Growth loop diagnosis	Diagnosing growth loops and selecting the next experiment	Decision memo
Social content fact-check rewrite	Reviewing and rewriting social content before publication	Claim table + draft
Growth launch readiness	Chaining context, creative, launch, analytics, approval, and learning checks	Launch packet
Learning extractor	Turning hard-won fixes into reusable knowledge	Lesson or skill proposal
Research-to-decision pipeline	Research that must become a decision	Decision memo
Repo triage workflow	Auditing unfamiliar codebases	Triage report

Example workflows

Example	Use it for
Fictional product audit	Combining research, review, approval, and learning workflows
Fictional market research	Turning public research into a decision artifact
Fictional repo triage	Mapping a repository before edits
MDX transcreation	Localizing MDX while preserving structure, links, code, and voice

Why this is different

Most public AI repositories show prompts, vendor skills, or tool recipes. This repo shows the operating layer around AI work: delegation contracts, approval gates, verification standards, learning loops, and measurable artifacts.

The goal is to make AI work safer and more legible, especially when agents touch real projects, code, tasks, or external systems.

Safety posture

All examples are synthetic. Do not commit secrets, private conversations, client/employer material, real account IDs, internal repo names, or hidden system prompts. See Publication policy.

Before release, run:

bun run validate

This checks tests, workflow schema compatibility, skill structure, README, examples index, eval fixture discoverability, eval fixture shape, and obvious publication-safety patterns. Use release-checklist.md for the full public release gate.

Accessibility

Accessibility expectations for docs, templates, diagrams, and CLI output are in ACCESSIBILITY.md.

Repository layout

cli/         Bun CLI for workflow validation, runbooks, audits, and scaffolds
principles/   Core operating beliefs
schema/      Machine-readable workflow schema
skills/      Public-safe skill drafts for repeatable AI-assisted work
workflows/   Reusable playbooks and executable-style workflow files
templates/   Copy/paste templates, AGENTS.md template, skill draft template
examples/     Synthetic case studies and eval prompts
diagrams/     Visual explanations

Contributing

Contributions should improve public-safe workflows, templates, examples, or safety practices. See Contributing.

License

CC BY 4.0. See License.