Skip to content

JPeetz/agent-skills / skills/business-strategy/product-manager-toolkit

product-manager-toolkit

Verified

Use this skill when performing product management tasks like prioritizing features, synthesizing customer interviews, writing PRDs, setting OKRs, or managing stakeholder communication. Activates on "product management", "PM toolkit", "prioritization", "RICE", "PRD", "feature prioritization", "customer interview", "OKR planning", "stakeholder update", or "roadmap planning". Covers RICE prioritization, customer interview synthesis, PRD authoring, OKR setting with leading/lagging indicators, stakeholder communication templates, and decision frameworks.

FileKindSize
CHANGELOG.mdmarkdown2.0 KB
LICENSEother1.1 KB
SKILL.mdskill md17.2 KB
evals/evals.jsonconfig8.0 KB
references/sources.mdmarkdown5.4 KB
scripts/rice_scorer.pyscript8.2 KB

evals/evals.json

View on GitHub
[
  {
    "name": "rice-ranking-with-override",
    "description": "Test that RICE ranking uses absolute reach, one effort unit and defensible confidence, that the script is actually run, and that a strategic override is recorded with a reason rather than silently reordering the list.",
    "test_input": "features.json has 7 candidate features for 2026-Q1 with reach as account counts per quarter, impact on the canonical 3/2/1/0.5/0.25 scale, confidence as a decimal, and effort in engineer-weeks. Two of them are 'SSO/SAML' (reach 90, impact 3, confidence 0.9, effort 12) and 'Bulk CSV import' (reach 1400, impact 2, confidence 0.8, effort 6). Run scripts/rice_scorer.py and tell me what to build first.",
    "expected_output": "A ranked markdown table from the script with reach, impact, confidence, effort and score visible per row, Bulk CSV import ranked above SSO/SAML on score, and an explicit statement that SSO is worth keeping despite a lower score because enterprise deals require it — recorded as an override with its reason, not as a re-scored item.",
    "assertions": [
      "scripts/rice_scorer.py is executed rather than the scores being estimated by hand",
      "Every row shows all four inputs, not just the final score",
      "Reach is reported as an absolute count (accounts per quarter), never a 1-5 rating",
      "Bulk CSV import outranks SSO/SAML on RICE score",
      "The SSO recommendation is preserved via a labeled override that names strategy or dependency as the reason",
      "The output states that RICE ranks but does not decide"
    ]
  },
  {
    "name": "interview-synthesis-evidence-discipline",
    "description": "Test that raw interview notes become behavior-named themes with participant counts, verbatim evidence and a counter-evidence section — not a narrative summary of the most quotable users.",
    "test_input": "Here are 14 customer interview transcripts from a B2B analytics tool (mixed enterprise and mid-market, 2 churned users). Synthesize what we learned and tell me what to build.",
    "expected_output": "A TL;DR of three sentences, then themes ranked by participants x severity where each theme is named by the behavior (e.g. 'rebuilds the weekly export in Excel every Monday'), carries a distinct-participant count of at least 3, names the segments present, quotes verbatim evidence with participant IDs, and states the current workaround. Followed by signals below the 3-participant threshold, a counter-evidence section, and open questions. No feature recommendations until the themes are established.",
    "assertions": [
      "Output opens with a three-sentence TL;DR before any theme detail",
      "Each theme names a behavior rather than a topic noun",
      "Each theme reports a count of distinct participants and the segments represented",
      "Themes below the 3-participant threshold are listed separately as signals, not promoted to themes",
      "Verbatim quotes are attributed to participant IDs",
      "A counter-evidence section exists and contains at least one item",
      "Current workaround is stated per theme",
      "The two churned users are distinguishable in the segment breakdown"
    ]
  },
  {
    "name": "prd-scope-and-kill-criteria",
    "description": "Test that a PRD request produces a bounded scope contract with must/should/won't split, one primary metric with baseline and window, and kill criteria — and that a vagueness in the brief is surfaced as an open question rather than invented away.",
    "test_input": "Write a PRD for 'make onboarding faster' for our project-management SaaS. We have 3,400 teams, activation (first shared project within 7 days) is 34%, and the exec wants it shipped this quarter.",
    "expected_output": "A 2-4 page PRD with a problem statement citing the 34% baseline, a one-sentence vision, at most two personas, a must/should/won't-have split where the won't-have list is named explicitly, requirements written as testable acceptance criteria, one primary metric (activation 34% -> a stated target by a stated date) plus guardrails, kill criteria written before launch, and an open-questions list that flags 'faster' and the exec's quarter deadline as decisions needing an owner.",
    "assertions": [
      "A 'Won't have this time' list is present and names at least one adjacent thing stakeholders would assume is included",
      "Requirements are testable and follow an as-a/I-can/so-that plus acceptance form",
      "Exactly one primary metric is named, with the 34% baseline, a target and a measurement window",
      "Kill criteria are stated before any launch or rollout plan",
      "At most two personas are defined",
      "Open questions carry an owner and a decide-by date",
      "The ambiguous term 'faster' is flagged rather than silently interpreted"
    ]
  },
  {
    "name": "okr-output-versus-outcome",
    "description": "Test that a draft OKR set containing output-style key results is diagnosed and rewritten into leading/lagging outcome KRs with baselines, and that confidence is rated at the start of the quarter.",
    "test_input": "Here are our draft Q1 OKRs for the platform team: Objective 'Be the most reliable platform in the category.' KRs: (1) Ship the new deployment pipeline by March 15, (2) Launch the status page redesign, (3) Complete the observability migration. Fix these.",
    "expected_output": "A diagnosis that all three KRs are outputs/tasks with no baseline, then a rewritten set of 2-4 outcome KRs each with baseline, target and date, at least one lagging KR (e.g. p95 incident recovery time, error budget burn) and one or more leading KRs, a high/medium/low confidence rating per KR with the assumption that would break it, and a note on how the shipped work becomes an input rather than a key result.",
    "assertions": [
      "All three original KRs are identified as output/task statements",
      "Every rewritten KR has a numeric baseline and a target with a date",
      "At least one lagging KR is present and labeled as lagging",
      "At least one leading KR is present and labeled as leading",
      "Each KR carries a high/medium/low confidence rating made at the start of the quarter",
      "The output states that shipping the pipeline is an input to a KR, not a KR itself",
      "Scoring guidance (0-1 scale, ~0.7 is a good stretch quarter) is stated"
    ]
  },
  {
    "name": "negative-launch-plan-out-of-scope",
    "description": "Test that a launch/channel request is routed to go-to-market-launch rather than answered with PM toolkit content.",
    "test_input": "Plan our Product Hunt launch for next Tuesday — tagline, gallery frames, first comment, and the day-of schedule.",
    "expected_output": "A short routing response identifying this as a go-to-market / launch-execution task and pointing to the go-to-market-launch skill, without producing a PH runbook from this skill.",
    "assertions": [
      "The response identifies the request as out of scope for product-manager-toolkit",
      "go-to-market-launch is named as the correct skill",
      "No Product Hunt runbook, tagline or gallery plan is produced",
      "The boundary is stated in terms of what this skill does cover (prioritization, PRDs, OKRs, synthesis)"
    ]
  },
  {
    "name": "negative-pricing-decision-out-of-scope",
    "description": "Test that a price/packaging question is not answered with product management frameworks, and that the skill does not invent a price.",
    "test_input": "We're launching a Pro tier. What should it cost, and which features go in it versus the free plan?",
    "expected_output": "A routing response identifying pricing and packaging as out of scope and naming pricing-strategist as the right skill. No price, no tier feature assignment, no willingness-to-pay figures invented.",
    "assertions": [
      "The response states that pricing and packaging are out of scope for this skill",
      "pricing-strategist is named as the correct skill",
      "No currency figure or price point is proposed",
      "No free-vs-Pro feature split is invented",
      "The response may note that the feature inventory itself can be produced here, but the packaging decision cannot"
    ]
  }
]
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall