The CI system went from a monolithic skill to a modular, structured intelligence system where context compounds across every mode.

What changed

Modular architecture

• Split /analyze-competitors from one 205-line file into a router (38 lines) + 6 workflow modules (landscape, deep-dive, synthesis, monitoring, guide, quality-rules)
• Each module loads on demand. Router detects the mode, reads the right file.

Upgraded landscape

• 4 competitor types (Direct, Indirect, Substitute, Potential) with identification techniques per type
• Structured competitor profiles with fixed fields
• Source and confidence tagging on every data point
• Positioning map with axis selection guidance
• Coverage assessment flagging what's verified vs shallow
• New template (template/competitive-landscape.md) with 7 sections

Upgraded deep dive

• Profile Summary at the top (structured fields that landscape uses for pre-population)
• Source and confidence tagging per claim (High/Med/Low)
• Competitor type classification (D/I/S/P) in frontmatter
• Seamless data flow into landscape universe table

Upgraded synthesis

• Evidence base structured around 6 categories: problem space, who's affected, competitive alternatives, our advantage, timing, risks
• Top 3 opportunities as "we should" statements grounded in evidence
• Coverage assessment: what's solid, shallow, missing, recommended next steps
• Owns the Strategic Insights section of context/competitors.md

Smart entry point

• Run /analyze-competitors with no arguments to see where you are and what to do next
• Reads context state, assesses CI maturity, recommends the specific next step
• Adapts to context: empty → onboard, no landscape → map the market, deep dives exist → synthesize

Onboarding upgrade

• Market section added to company.md (market definition, JTBD, ICP segments, scope)
• Onboarding flow reordered: company → market → competitors → personas → product
• Market questions added to follow-up step

How it connects

Context IS the chain. Each mode reads what others wrote to context/competitors.md:

• Deep dives feed landscape (pre-population via Profile Summary)
• Landscape feeds synthesis (market structure + profiles)
• Synthesis produces Strategic Insights section
• Monitoring tracks drift
• No orchestration engine needed

What's next

• Opportunity Assessment using Cagan's POA (POD-31)
• Strategic Critic agent (POD-32)
• Context file templates (POD-34)
• Automated monitoring via /schedule (POD-20)

What's new

Contributor infrastructure. PM Pilot is now ready for outside contributions.

Added

• CONTRIBUTING.md — how to add skills, templates, and improvements. Covers repo structure, skill anatomy, testing guide, versioning rules, PR process, and product data policy.
• GitHub issue templates — bug report, feature request, and skill feedback. Blank issues disabled; contributors are guided to the right template.
• Issue template config — links to Discussions and email for questions.

Changed

• README rewritten — "Why This, Not a Blank AI Chat" positioning, ASCII workflow diagram showing how skills chain, context system table, prerequisites, contributing link, license.
• README version updated to v1.5.0 (now v1.6.0).
• AGENTS.md — added missing /product-teardown to command list.
• LICENSE — updated copyright to full name (Arezou Solouki).

Fixed

• .gitignore — added context/*/ to prevent accidentally committing deep dives, teardowns, and other context subfolders.
• Removed obsolete setup/ directory — leftover from early development, no longer needed.
• Removed unused VS Code workspace file (pmos-starterkit.code-workspace).
• Committed investor-update.md style template that was sitting untracked.

What's new since v1.3.0

/product-teardown skill (v1.4.0)

New skill for tearing down competitor products: features, UX flows, architecture, and growth mechanics. Complements /analyze-competitors deep dives with product-level analysis. Three depth levels (light/standard/deep). JTBD-per-feature analysis, user journey walkthrough, and synthesis with steal/counter/ignore recommendations.

/analyze-competitors v1.5.0

Synthesis is its own mode. No longer silently activated inside landscape. Distinct trigger phrases, inventory step with staleness checks, cross-cutting analysis (convergence, divergence, gaps, threat ranking) instead of per-competitor summaries. Flags missing teardowns instead of guessing.

Audience and tone. Output adapts when the PM says who it's for: exec (strategic, concise), sales (win/loss, objections), engineering (architecture, tech stack). Defaults to PM-level detail.

Edge cases handled. Empty-directory fallbacks for internal intel (deep dive, landscape, monitoring). Duplicate detection asks before creating a second deep dive. Paywalled source guidance for G2/Capterra/Gartner.

Cleaner structure. Four top-level workflow sections (Landscape, Synthesis, Deep dive, Monitoring) replace nested subsections. Dropped redundant frontmatter status field.

Competitive landscape synthesis

/analyze-competitors landscape mode now auto-detects when 3+ deep dives or teardowns exist and offers synthesis vs fresh research. Synthesis reads all existing analysis and produces cross-competitor patterns, positioning map, capability landscape, and strategic bets.

New templates

• Product teardown template with frontmatter (product, company, persona, goal, depth level)
• Competitive landscape template redesigned with market overview, direct vs adjacent competitors, capability & UX landscape, pricing/GTM patterns, and implications sections

Skill READMEs complete

All 12 skills now have README.md files with usage examples and tips.

Full Changelog: v1.3.0...v1.5.0

What's new

/critique skill

New skill that pressure-tests documents (PRDs, strategy, pitch, positioning) against your product context. Finds gaps, weak logic, and unvalidated assumptions.

Codex support

Added Codex entry point and README guidance for using PM Pilot with OpenAI Codex. (Thanks @joneri — #2)

Structured competitive deep dive template

Replaced the lightweight freeform template with a structured format:

• Frontmatter metadata: name, segment, icp, overlap_area, threat_level, status, last_reviewed
• 6 sections: Snapshot, Product & Experience, Pricing, Strategy, Evidence, Implications
• Feature matrix (vs us) for direct comparison
• Concrete actions split into copy/adapt, counter-position, ignore/de-prioritize
• Dual status: frontmatter tracks research completeness, body tracks document lifecycle

Other changes

• Added version field to all skill files
• /analyze-competitors bumped to v1.1.0 with updated deep dive workflow

Full Changelog: v1.0...v1.3.0

PM Pilot: AI copilot for product managers. Built for Claude Code and Cursor.

What's included

9 skills covering the core PM loop:

• /prd - Write PRDs, one-pagers, project briefs
• /analyze-competitors - Deep dives and landscape analyses
• /synthesize-interviews - Turn interview notes into themes and recommendations
• /meeting-notes - Transform transcripts into decisions and action items
• /status-update - Status updates matched to the audience
• /prioritize - Score and rank features against real context
• /triage-feedback - Categorize, prioritize, and route customer feedback
• /break-down - Break features into buildable work items
• /release-notes - Draft user-facing release notes

Context-first architecture: Build knowledge about your company, competitors, personas, and product. Every skill draws from that context.

Flexible input: Paste content, drop a file path, or reference workspace files. No rigid setup required.

Getting started

1. Clone the repo
2. Open in Claude Code or Cursor
3. Run /onboard or just start talking

Feedback welcome: open an issue or start a discussion.