Action approval planner

Skill duandigi/duandigi-growth-marketing-skill/skills/action-approval-planner

Evidence-first AI Agent Skills for growth marketing — multi-channel analytics (SEO, paid media, social, local search, CRM), secure account integration, AI evaluation, experimentation & approval-safe optimization for Claude Code.

Install

npx -y skills add duandigi/duandigi-growth-marketing-skill --skill action-approval-planner

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

25 days oldThe repository was created 25 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use this skill when converting an AI recommendation into a safe observe, investigate, prepare, approve, execute, verify, or rollback workflow for marketing, website, CRM, advertising, content, or data actions.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

3.6 KB, as published. Nobody here has run it

Action Approval Planner

Purpose

Ensure AI-generated optimization is reviewable, reversible, authorized, and matched to the financial, privacy, platform, brand, and production risk of the action.

Inputs

Recommended action, affected provider, project, asset, and business objective
Current connection capabilities and application roles
Expected impact, cost, privacy, policy, brand, and production risk
Backup, draft, feature flag, test, logging, verification, and rollback options

If a required input is unavailable, label it unknown, state how it limits the decision, and create a collection, mapping, validation, or instrumentation task. Never invent credentials, assets, metrics, permissions, or business outcomes.

Workflow

Classify the action as observe, investigate, prepare, execute, publish, spend, contact, delete, access, or irreversible.
Determine the minimum provider permission and application role required.
Assign an approval tier based on money, audience, data, access, production, policy, and reversibility.
Require a preview, diff, draft, simulation, test environment, or dry run whenever feasible.
Define preconditions, approver, expiry, execution window, idempotency, and rollback.
Log the request, evidence, decision, executor, result, verification, and residual risk.
Return a machine-readable approval card and human summary.

Required output

Return a concise, decision-oriented result containing:

Action classification and approval tier
Required capability, role, and approver
Preview, preconditions, tests, and execution plan
Verification, audit log, and rollback plan
Expiry, stop conditions, and residual risk

Label material statements as confirmed, calculated, inferred, assumed, or unknown. Include the data period, last complete period, source lineage, and confidence whenever they can change the decision.

Guardrails

Never execute spend, publishing, bulk contact, deletion, access, or irreversible changes without explicit authorization.
Do not let an AI agent approve its own high-impact action.
Do not reuse an approval for a materially different scope, asset, budget, audience, or time period.
Do not execute when backup, verification, or rollback requirements are unmet.
Do not claim guaranteed growth or present an estimate as observed fact.
Do not reveal secrets, personal data, private provider payloads, or cross-project information.
When an action can spend money, publish, contact people, alter access, modify production, or delete data, prepare an approval request instead of executing automatically.

Completion check

Before finishing, verify that the output:

answers a specific business or implementation decision;
uses the correct organization, project, asset, date range, time zone, and currency;
separates performance problems from data, connection, and attribution problems;
includes evidence, uncertainty, affected scope, and a measurable next step;
respects least privilege, approval, audit, and rollback requirements;
is no longer than necessary for the decision.