product-skills
by alirezarezvaniproduct-skills is a Product Management orchestrator that routes discovery, prioritization, analytics, roadmaps, PRDs, experiments, and AI evals to the right lane, with scripts for OST linting and discovery cadence checks.
This skill scores 84/100, which means it is a solid listing candidate for directory users who want an agent to coordinate product-management work rather than rely on a generic prompt. The evidence shows clear activation cues, substantial workflow content, deterministic scripts for routing and validation, and enough reference material to support credible product discovery and planning loops. The main adoption caveats are packaging clarity and the complexity of depending on a larger product-skill ecosystem.
- Strong triggerability: the frontmatter names concrete user intents such as prioritization, product experiments, discovery loops, and OST validation, and distinguishes itself from project-management, marketing, and engineering skills.
- Good agent leverage through deterministic support scripts: the product goal router, discovery cadence tracker, and OST linter provide machine-checkable routing and quality gates rather than only prose advice.
- Useful supporting references and samples: the repository includes discovery-log and OST JSON examples plus canon documents for AI product evals, continuous discovery, and product operating models.
- No install command or README is present in the skill path, so users must infer installation from the broader repository/tooling conventions.
- It is an orchestrator for 16 product-team lanes; users who only need one narrow PM workflow may find the routing layer heavier than a standalone skill.
Overview of product-skills skill
What product-skills is for
product-skills is a Product Management orchestrator skill for routing messy product work to the right product sub-skill, then keeping the work tied to outcomes, discovery evidence, and measurable decision gates. It is best suited for PMs, product leads, founders, product trios, and AI-assisted product teams who want help with questions like “what should we build?”, “how should we prioritize?”, “is this Opportunity Solution Tree sound?”, or “how do we turn discovery into an experiment plan?”
Best-fit Product Management use cases
Use the product-skills skill when your request spans product strategy, discovery, analytics, prioritization, roadmapping, UX research, experiments, user stories, PRDs, AI product evals, or product operating model decisions. Its main value is not writing a prettier generic PM document; it helps classify the work, choose one lane, and apply product-specific constraints such as outcomes over outputs, assumption testing, OST structure, and measurable quality gates.
What makes it different from a generic prompt
The skill includes a deterministic router, scripts/product_goal_router.py, plus two practical gates: scripts/discovery_cadence_tracker.py for continuous discovery health and scripts/ost_linter.py for Opportunity Solution Tree structure. That means product-skills can do more than advise: it can check whether a discovery habit is healthy, whether an OST has feature-shaped “opportunities,” and whether targeted opportunities have multiple solutions and tests.
When product-skills is the wrong fit
Do not install product-skills if you only need project delivery tracking, sprint coordination, or engineering task execution. The repository explicitly separates product direction from project management and generic agent loops. It is also not a replacement for customer evidence, analytics access, or stakeholder judgment; it works best when you provide real goals, metrics, interviews, assumptions, and constraints.
How to Use product-skills skill
product-skills install and first files to inspect
Install from the parent GitHub skills repository with:
npx skills add alirezarezvani/claude-skills --skill product-skills
Then inspect the skill folder at product-team/skills/product-skills. Read SKILL.md first to understand triggers, routing, and hard rules. Then review references/product_operating_model.md, references/continuous_discovery_canon.md, and references/ai_product_evals.md to see the product logic behind the prompts. If you plan to use the scripts, open assets/sample_discovery_log.json and assets/sample_ost.json before adapting your own data.
Inputs that make product-skills usage work well
The skill performs best when your prompt includes five things: the product outcome, current metric baseline, target user or segment, evidence available, and the decision you need. Weak input is “help me prioritize features.” Strong input is: “Use product-skills for Product Management. We are a B2B SaaS onboarding team. Outcome: raise week-4 retention from 22% to 30%. Evidence: 8 support tickets about setup uncertainty, 5 interviews, activation funnel drop at integration verification. Options: checklist, sample-data preview, concierge setup. Help choose the right lane and produce the next decision artifact.”
Suggested workflow for a real product decision
Start with a rough product question and ask the agent to route it using product-skills rather than forcing a framework yourself. If the issue is unclear, ask for clarifying questions before output. For discovery-heavy work, create or adapt an OST JSON and run ost_linter.py as a structural check. For continuous discovery, maintain a discovery log and run discovery_cadence_tracker.py weekly. For AI features, use references/ai_product_evals.md to require a golden set, rubric, and guardrail metrics before treating a PRD as complete.
Practical command-line checks
The Python scripts use standard library patterns and are designed for deterministic gates. Typical usage is to save your JSON input, then run:
python scripts/ost_linter.py path/to/ost.json
python scripts/discovery_cadence_tracker.py path/to/discovery_log.json
Use the exit codes as workflow signals. For example, an OST linter failure should block a roadmap that cites the tree, because the tree may contain orphan solutions, feature-phrased opportunities, or untested solution ideas.
product-skills skill FAQ
Is product-skills only for senior PMs?
No, but beginners need to provide concrete context. The product-skills guide can help newer PMs avoid common traps such as roadmaps full of output commitments, OKRs that are actually shipping lists, and discovery that is not tied to an outcome. Senior PMs will get more value from the routing, eval, and gating patterns because they can plug them into an existing operating cadence.
How is product-skills different from RICE or OKR templates?
RICE and OKRs are lanes inside the broader product-skills system, not the whole system. The orchestrator is useful when you are unsure whether the right next step is prioritization, discovery, analytics, experiment design, roadmap framing, PRD work, or AI evaluation design. It helps avoid applying RICE to every product decision when the real gap may be weak evidence or an invalid outcome.
Can product-skills work outside Claude Code?
The skill metadata lists compatibility with tools such as claude-code, codex-cli, cursor, antigravity, opencode, and gemini-cli. In practice, the markdown guidance is portable, and the Python scripts can be run wherever Python is available. Installation and invocation details may vary by host tool, so confirm how your agent platform loads external skills.
What should I prepare before installing?
Prepare at least one real product problem, a metric or desired outcome, and any available customer or analytics evidence. If you want to test the deterministic parts, prepare a discovery log matching assets/sample_discovery_log.json or an OST matching assets/sample_ost.json. Without real inputs, the skill will still generate structure, but its recommendations will be less trustworthy.
How to Improve product-skills skill
Improve product-skills results with sharper prompts
Name the desired artifact and the decision it must support. Instead of asking “write a PRD,” ask: “Route this with product-skills, identify whether the missing work is discovery, analytics, prioritization, or spec writing, then produce only the artifact needed to decide the next step.” This prevents premature document generation and lets the router select the right product lane.
Add evidence before asking for recommendations
The biggest failure mode is asking for confident product strategy from thin context. Add interview excerpts, funnel numbers, segment differences, support themes, experiment history, and known constraints. For AI product work, include sample inputs, bad outputs, safety concerns, and quality expectations so the answer can include eval design rather than vague acceptance criteria.
Iterate after the first output
Treat the first answer as a diagnostic draft. Ask which assumptions are unsupported, which metric would change the recommendation, and what evidence would invalidate the plan. If the output proposes a solution too early, push it back through an OST: outcome, opportunities, multiple solutions, and assumption tests. Re-run the linter or cadence tracker after changes.
Common failure modes to watch
Watch for feature-shaped opportunities, single-solution thinking, vanity metrics, discovery logs with no weekly rhythm, and AI feature PRDs without golden sets or rubrics. product-skills is strongest when you let its constraints block weak product work instead of using it only to produce polished PM prose.