feat: add pm-data skill — quantitative analytics layer with PostHog, Amplitude, and pipeline integration#7
Open
alexhumeau wants to merge 5 commits into
Open
Conversation
…Calendar and Granola connectors New skills: - pm-standup: daily briefing from git, Linear, Google Calendar, and Granola Surfaces yesterday's commits, today's meetings with prep gaps flagged, top priorities inferred from roadmap, blockers, and drift detection - pm-interview: user interview guide + Granola-powered signal extraction Prepares structured interview guides, detects Granola transcripts automatically, extracts findings into FEEDBACK.md (consumed by /pm-audit and /pm-user-feedback) - pm-weekly-update: stakeholder update email adapted to audience Supports manager, CEO/exec, investor, dev team, and cross-functional formats New connectors: - google-calendar: MCP-based (mcp__claude_ai_Google_Calendar__*), fetches today's events - granola: MCP-based (mcp__claude_ai_Granola__*), fetches meeting notes and transcripts pm-standup integrates both connectors: shows today's schedule, flags meetings with no recent commits (prep gap), and surfaces user calls for /pm-interview. pm-interview checks Granola for an existing transcript before asking for manual notes. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…t OS usage When running from a Product OS folder (not inside a codebase), pm-standup now fetches commits from all active GitHub repos instead of reading local git. Commits are grouped by repo: [repo-name] commit message. On first run, caches the list of repos with recent activity to speed up subsequent runs. Falls back to local git if GitHub is not connected. This makes pm-standup useful as a single morning command regardless of which repo you're currently in. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…a, Alvarez, NNG
Complete rewrite of the interview guide generation, grounded in four recognized
frameworks:
- Teresa Torres (Continuous Discovery Habits): one story per interview, temporal
anchoring, excavation of the timeline with neutral prompts, anti-generalization
redirect ("usually" → "last specific time")
- Rob Fitzpatrick (The Mom Test): past behavior only, no hypotheticals, mandatory
workaround/implication/alternative questions, explicit anti-patterns table
- Bob Moesta (JTBD Switch Interview): 6-stage buying timeline for churned/switched
users, four forces mapping (push/pull/anxiety/inertia), contrast probing technique
- Cindy Alvarez (Lean Customer Dev): hypothesis-first ordering (Constable), top 3
questions before the call, forward action at close, post-call review discipline
Session type auto-detection: current user → story-based, churned → JTBD switch,
prospect → lean discovery, buyer → positioning framing.
Signal reliability thresholds now explicit: 3 min, 5 standard, 8-12 for full JTBD map.
Four forces map added to FEEDBACK.md output structure.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
New skill /pm-data answers a specific product question with quantitative data. Supports: trend analysis, funnel drop-off, retention (D1/D7/D30), user paths, cohort comparison, and feature impact measurement. Data sources (in priority order): - PostHog MCP (mcp__claude_ai_PostHog__*) — full support: trends, funnels, retention, paths, stickiness, custom HogQL queries, event definitions - PostHog API (POSTHOG_API_KEY + POSTHOG_PROJECT_ID) - Amplitude API (AMPLITUDE_API_KEY + AMPLITUDE_SECRET_KEY) - Manual paste fallback (always works) Every finding includes a confidence level (high/medium/low) and a "so what" interpretation — no raw number dumps. Triangulates with FEEDBACK.md qualitative signal when available. Output written to .nanopm/DATA.md (append — preserves prior analyses). DATA.md is consumed by /pm-audit and /pm-prd for quantitative backing. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
pm-audit (Phase 2 + Phase 4 + AUDIT.md template): - Reads DATA.md if present, extracts high-confidence metrics and flagged unknowns - Phase 4 synthesis now cross-references quanti vs quali signal — contradictions between DATA.md and FEEDBACK.md are flagged explicitly as the most valuable findings - New Section 5 in AUDIT.md: "What The Data Says" (omitted if no DATA.md) pm-prd (Phase 2 + Problem Statement template): - Reads DATA.md for funnel/retention metrics relevant to the feature being specced - Problem Statement template now includes a slot for one quantified fact from DATA.md - Success Criteria baseline targets can be derived from DATA.md metrics - Only 🟢 high-confidence findings are used — low-confidence data is not cited as fact DATA.md is now a first-class input to the nanopm pipeline alongside FEEDBACK.md. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What this adds
nanopm's pipeline is entirely qualitative. This PR adds the missing quantitative layer: a dedicated skill to answer product questions with data, integrated into the existing
pm-auditandpm-prdpipeline.New skill:
/pm-dataAnswer a specific product question with data. Enforces one question per run — vague questions produce useless output.
Supported analysis types:
Every finding includes a confidence level 🟢🟡🔴 and a plain-English "so what". No raw number dumps.
Output:
.nanopm/DATA.md(append — preserves history across runs).New connectors
posthogmcp__claude_ai_PostHog__*POSTHOG_API_KEY+POSTHOG_PROJECT_IDamplitudeAMPLITUDE_API_KEY+AMPLITUDE_SECRET_KEYBoth degrade to manual paste if unavailable.
Pipeline integration
DATA.md is now a first-class input alongside FEEDBACK.md.
/pm-auditchanges (minimal, surgical):/pm-prdchanges (minimal, surgical):Quanti / quali triangulation
If FEEDBACK.md exists,
/pm-datachecks for qualitative signal that confirms or contradicts the numbers. A funnel drop-off confirmed by interview quotes → strong signal. A number with no qualitative backing → flagged as hypothesis.Design principles