$ show deanpeters/product-manager-prompts
This repository is a comprehensively governed, production-grade library of 115+ prompt templates, meta-prompt generators, workshop guides, and structured routines for AI-assisted product management. It is not a deployed software application but rather a content methodology repository with application-like engineering discipline: standalone prompt contracts, coupling levels, CI-validated metadata, deterministic catalog generation, and an explicit coupling remediation plan. The library covers the full product management lifecycle—discovery, strategy, requirements, competitive intelligence, end-of-life, storytelling, and team facilitation—using three defined interaction modes (facilitation, co-construction, autonomous investigation). Notable sub-collections include an eight-discipline market intelligence framework (OSINT, FININT, GEOINT, etc.) with an all-source fusion prompt, a complete product sunset (EOL) workflow, and a TAM-SAM-SOM market sizing Langflow pipeline. The repo exhibits hybrid character: while primarily a content asset, it includes reusable Python scripts for validation and catalog generation, a GitHub Actions CI pipeline, and follows software engineering best practices such as versioning, changelogs, and architectural decision records.
The LLM identified 5 files worth highlighting for reuse:
Enforces structural standards across all prompt files: metadata blocks (Description, Usage Note, etc.), coupling levels, standalone declarations, skills contract compliance, and licensing sweep. Reusable for any markdown-based content library that needs automated quality gates. Demonstrates thoughtful coupling detection (forward pointers allowed, backward prerequisites flagged as debt) and integrates with GitHub Actions.
Auto-generates a flat index of all prompts from metadata comment blocks, producing a deterministic YAML output. Enables a CI check that fails if the generated catalog is stale. Reusable for any structured content repository where a machine-readable catalog is needed from annotated markdown files.
Defines the canonical v2 facilitation pattern with standing bypasses (best guess, bulk drop), loop control (skip, go back, stop early), context-aware option ranking, and question budgets. Used across most prompt-generators. Reusable as a design pattern for building interactive, multi-turn AI assistants that need to guide users through structured decision-making.
Consolidates findings from multiple intelligence disciplines into a single fusion brief. Enforces confidence stacking: 1 discipline = watch item, 2 = working hypothesis, 3+ = actionable, conflict = bluffing indicator. Reusable for any multi-source research synthesis task where evidence labeling (Fact/Inference/Assumption) and source citation are required.
Example skill packaged per Anthropic Agent Skills spec with YAML frontmatter (name, description, license, allowed-tools), template file, and skill metadata. Demonstrates the skill packaging standard that could be reused for distributing AI agent capabilities across platforms. Isolated and self-contained with no internal dependencies.