AI-assisted prompts, workflows, and examples for TPM productivity and program management. Built from real use - things that save time on actual work, not proof-of-concept demos.
The tools here are not about replacing judgment. They're about removing mechanical work so there's more time for the work that actually requires it.
Each workflow follows the same pattern: a context or persona file that carries stable domain knowledge, and a prompt file that drives the session interaction. Keeping them separate means you can reuse the domain knowledge across different prompt formats, and swap the prompt without rewriting the expertise.
| Workflow | What It Does |
|---|---|
| Status Report Generator | Reads a folder of raw inputs - meeting notes, Slack exports, RAID log updates, ticket summaries - and produces a structured draft status report. Nine sections including overall status, decisions needed, risks and issues, and a NEEDS REVIEW section that flags what requires TPM judgment before sending. Integrates with the Slack Crawler in tpm-toolbox. |
| Landscape Planner | AI landscape architect consultation for Sonoma County conditions. Give it your budget, yard situation, and goals. Receive a visual layout, phased implementation guide, plant shopping list, and budget breakdown. Covers fire safety, drought tolerance, deer and gopher pressure, and local nursery recommendations. |
| App Builder | Two-prompt system for going from a rough app idea to a working web application. The intake prompt interviews a non-technical person and produces a structured developer brief. The builder prompt takes that brief through six phases - idea refinement, tech stack, architecture, MVP scope, UI/UX direction, and feature-by-feature build with complete copy-paste-ready code. |
The prompts and workflows above answer "how do I do this piece of work well." The frameworks answer the question underneath that one: which work should AI be touching at all, who does it, and where does the human stay.
| Framework | Answers |
|---|---|
| AI Execution Tiers | How much AI belongs in a given activity, on a five-tier scale from human-only to AI-only, and how to classify without kidding yourself. |
| Activity Mapping Method | How to map an entire company's work against those tiers in about a week, at three levels of resolution. |
| Multi-Agent Delivery Framework | 19 roles across 8 phases, ideation through operations, with senior/junior tier selection logic and an audit trail. Describes the framework; the individual role specifications are not published. |
| AI Tool Decision Matrix | What to buy and in what order, and why the AI layer is phase 4 of 5 rather than the first purchase. |
Regulated environments change the adoption order. That overlay lives in security-program-playbooks.
These four were listed here as "coming" for a while. Three of them had already shipped into this repo and the list didn't keep up, which is a good argument for indexing what exists rather than advertising what might.
| Workflow | What It Does | Where |
|---|---|---|
| Meeting Notes Processor | Takes raw meeting notes and extracts decisions, action items, and open questions in structured format. Output slots into the meeting notes tracker in tpm-toolbox. | meeting-notes-to-actions.md · worked example: input / output |
| Pre-Meeting Brief Generator | Given an agenda and participant list, produces a one-page brief: the one outcome worth protecting, what each participant will push on, which decisions can actually close, and the objection you would least like to be asked. | pre-meeting-brief.md |
| Escalation Draft | Takes a problem description and produces a structured escalation: what is blocked, the impact, what has been tried, the decision needed, and the deadline. Hard to write under pressure, easier from a starting point. | escalation-draft.md · worked example: sanitized escalation |
| Compliance Intake Triage | Turns a regulatory requirement, audit finding, or control request into engineering work with owners, evidence needs, milestones, and risk visibility. First cut only, not a substitute for GRC review. | compliance-intake-workflow.md |
Nothing is queued beyond this right now. Additions happen when a prompt has been used on real work often enough to be worth writing down, not on a roadmap.
Output you can use, not output you've to rewrite. Prompts here are tuned for specific formats and real contexts. They produce things that slot into existing workflows, not generic text that needs heavy editing.
Transparent about what AI is good at and what it's not. AI is good at structure, first drafts, and pattern recognition. It's not good at knowing your specific situation, your org dynamics, or what the executive in the room actually cares about. These tools handle structural work. Judgment stays with you.
Documented well enough to adapt. Every workflow includes the prompt, expected input format, expected output format, and notes on where it tends to go wrong. If you need to adapt it to your context, you should be able to do that without reverse-engineering it.
Lightweight. No frameworks, no dependencies where avoidable. Prompts you can run in Claude.ai or via API. If something requires infrastructure, it is documented clearly and kept minimal.
- tpm-templates - The output formats the status report generator produces match the templates there
- program-reporting-frameworks - The status and steering committee frameworks explain what goes in each section and why
- tpm-toolbox - The Slack Crawler produces intelligence briefs that feed directly into the status report generator; the canvas synthesis tool handles the other direction
If you've a prompt, workflow, or pattern that saves real time on meaningful work and can be documented well enough to be useful to someone else, open a PR or file an issue. The bar: it has to work, it has to be documented, and it has to be general enough to apply outside one specific organization.
Built from experience running platform security, infrastructure, and compliance programs at enterprise scale. Maintained by Eric White | ChefPlex
Copyright (c) 2026 Eric White. Licensed under CC BY 4.0: use it, adapt it, put it to work in your own program. Credit is the only condition.