Icp

Pack Availability Guard

Before telling the user to run a skill from another project-local pack, check .agents/project.json.enabled_packs. If the target pack is not enabled, recommend $pack install <pack> instead of the target skill. Global skills are always valid. Skills from this same pack are valid because the current skill is already running from that pack.

ICP — Research-Driven Customer Discovery

Invoke as $icp.

Report-First Approval Gate

Default to report-only: present findings, evidence coverage, assumptions, recommended artifact path, and proposed file changes in a pre-approval alignment page plus a concise conversation summary for user approval before creating or updating canonical research, spec, or task files.

Do not write or overwrite synthesized deliverables until the user explicitly approves, unless the user invoked an explicit write/update/fix mode or clearly asked to write files upfront. Raw evidence capture may be persisted before analysis when reproducibility requires it; report those raw paths separately and still gate synthesized research/report writes.

When stopping for approval, build and attempt to open the alignment preview page first, then ask the user to review it and approve, question, or request adjustments. Do not include Recommended next skill, Recommended next command, or downstream routing language. The approval request itself is the next action. Only emit next-skill routing after the approved artifact has been written or updated.

Staged Research Workflow

Use this staged workflow for synthesized research or report outputs that would create or update canonical research, spec, or task files.

1. Stage 1 - Research and clarify. Perform the research, run required source/code checks, and ask any needed clarification questions. Write only a non-canonical working packet: flat mode uses research/_working/preliminary-<skill>-research.md; product-path mode uses research/{slug}/_working/preliminary-<skill>-research.md. Replace <skill> with this skill's name value. Do not create or update canonical research, spec, or task files in Stage 1. Raw evidence or search logs may remain as supporting evidence where this skill already requires them, but synthesized deliverables stay in the working packet.
2. Stage 2 - Review alignment. Consume the working packet and build the review HTML alignment page. The page must render the full preliminary packet, evidence matrix, assumptions/confidence register, source coverage gaps, proposed canonical file changes, and approval gates. Stop for either feedback-only YAML or final compiled YAML. Feedback-only YAML revises the working packet and page, then remains in Stage 2.
3. Stage 3 - Finalize approved artifacts. Consume final compiled YAML only when it has no unresolved needs-clarification, unresolved down feedback, or other unresolved negative feedback. Apply approved edits first, archive the working packet to docs/history/archive/YYYY-MM-DD/HHMMSS/<original-working-path>, remove the active working packet, write the approved canonical artifacts to the unchanged output paths below, and convert the alignment page to confirmed with the approval record preserved.

Canonical output paths remain unchanged. Search logs and other supporting evidence remain allowed only where this skill's output contract already requires them.

Automated research that identifies multiple ICP candidates, maps pain points and value props, and selects a primary ICP. Uses web search + codebase analysis instead of interviews. If research/idea-brief.md or research/{slug}/idea-brief.md exists, use it as starting context for the research frame.

Default stance: assume the user has no insider knowledge of the market. Explain segments, pain signals, and tradeoffs from first principles so the recommendation is defensible without founder intuition. Ask for corrections, hard constraints, or proprietary facts only when needed.

Evidence And Feedback Handling

Treat user feedback as input to evaluate, not as automatic ground truth.

• For factual, evidentiary, technical, or source-backed claims: verify against available evidence. If the user appears to misunderstand the evidence or states something factually incorrect, push back clearly and cite the evidence. Do not rewrite findings merely to agree.
• For taste, brand, positioning preference, risk appetite, prioritization, or other subjective judgment calls: weigh user feedback heavily and adapt the recommendation unless it conflicts with verified evidence.
• When feedback mixes facts and preference, separate them explicitly: correct the factual part, then incorporate the preference where it is a legitimate judgment call.
• When uncertain, say what is known, what is inferred, and what would change the conclusion.

Process

0. Product-Path Scope Resolution

Resolve research scope by product path before using code or app structure as a hint:

1. If $ARGUMENTS names a non-archived research/{slug}/ directory or a product-path ID whose scope_path points there, use that path. Treat {slug} as the product/app name, not the ICP, audience, or segment label.
2. If $ARGUMENTS names only research/_archive/{slug}/ or a manifest entry with status: archived or legacy status: abandoned, stop and warn that the path is archived; do not write or update scoped outputs there.
3. Read research/.progress.yaml when present. Normalize legacy active_path to active_paths on read and write back active_paths on manifest updates. Treat legacy abandoned as archived; exclude archived, abandoned, deferred, revisit_candidate, promoted, and any scope_path under research/_archive/ from active target selection.
4. If active product paths exist in the manifest, use those paths. If multiple active paths exist, ask which one to target unless this skill explicitly supports cross-path output.
5. If no active manifest target exists, list non-archived product directories under research/, excluding research/_archive/ and dot directories. Auto-select only when exactly one exists; ask when multiple exist.
6. If no product directories exist, use flat research/ single-product mode.
7. Detect monorepo/app/package structure only as a secondary hint. Suggest creating a missing research/{slug}/ product path when code clearly exposes an app, but do not require code or monorepo detection before using research/{slug}/.

When product path {slug} is active, read and write research under research/{slug}/, specs under specs/{slug}/, and treat top-level research/*.md files as flat-mode documents or cross-path summaries.

1. Parse input and gather concept context:
   • Read research/.progress.yaml when present. Normalize active_path (singular legacy) to active_paths (plural list) when reading; treat legacy abandoned as archived and exclude archived/deferred/revisit/promoted paths plus research/_archive/ scopes from active target selection. Use active_paths to identify current product/app/ICP focuses and product_paths[] to preserve secondary product paths without treating them as git branches or parallel implementation lanes.
   • Provisional product-path evidence. When a referenced product path is not present in research/.progress.yaml (either absent entirely or not in active_paths or product_paths[]), do not treat it as a canonical active path. Before using it as source context, require an explicit provisional-path evidence reference: a review-only-approved alignment page (e.g., alignment/idea-scope-brief-{topic}.html with approval_status: review-only-approved) that fully renders the proposed path's concept, brief, and manifest entry. If no such evidence exists, ask the user whether to proceed with the path as unverified context or to run $idea-scope-brief first.
   • Read $ARGUMENTS as a spec file path or concept text when provided.
   • Read research/{slug}/idea-brief.md in product-path scope, or research/idea-brief.md in flat scope, when present. Treat it as starting context and source hypotheses, not as settled truth.
   • If $ARGUMENTS is empty and an idea brief exists, use the idea brief as the primary input before falling back to README, specs, or codebase inference.
   • If $ARGUMENTS conflicts with the idea brief, flag the mismatch at the first checkpoint and ask which premise should guide ICP research.
   • Read codebase if it exists. Read existing research for background.
   • Use code/app hints secondarily (turbo.json, pnpm-workspace.yaml, lerna.json, nx.json, package.json workspaces, app folders, or package folders) only after product-path resolution. If multiple active product paths exist, or if this run is explicitly producing a cross-path overview, run the full ICP process per product path and produce research/{slug}/icp.md per path plus a unified research/icp.md. If code clearly exposes a user-facing app without a matching product path, suggest creating research/{slug}/ instead of treating monorepo detection as a gate.
   • Migrate flat files when product paths are introduced: If canonical flat research/*.md files exist and the user chooses or creates a product path, offer to move path-specific canonical docs into research/{chosen-slug}/. Leave or regenerate top-level files only when they are cross-path summaries.
   • Migrate old convention: If research/icp-{slug}.md files exist (old naming), offer to move them to research/{slug}/icp.md (and corresponding search logs to research/{slug}/icp-search-log.md). Create the subdirectories as needed.

Marketplace Side Preflight

Before identifying ICP candidates, resolve market-structure side coverage:

• Read any Market Structure Handoff or marketplace/platform/B2B2C/multi-sided notes from the idea brief, especially ## ICP Readiness. Treat side names and value exchange as hypotheses from the idea-scope brief, not validated ICPs.
• If no idea brief exists, infer likely sides from $ARGUMENTS, README, specs, codebase context, or product description when the concept suggests a marketplace/platform/B2B2C/multi-sided model.
• During broad market research, validate or refute the marketplace/platform/B2B2C/multi-sided classification with evidence; log the evidence for or against each apparent side and value exchange.
• Before candidate generation, write a side-coverage note in the working packet or search log that names material sides, buyer/customer/user distinctions, and any uncertain or unsupported side assumptions.
• Candidate generation must cover each material side or explicitly explain why a side is excluded, deferred, or not a customer side.

2. Broad market research: WebSearch with 8–12 query strategies (personas, pain points, segments, trends, competitors, forums, job postings, industry reports, business model, willingness-to-pay signals). Log all queries and findings. Classify the business model into one or more of: B2B SaaS (PLG), B2B SaaS (SLG), B2C, B2C subscription, marketplace/platform, B2B2C, D2C, open-source/open-core, API/developer-first. Document classification with evidence in search log — this gates which sub-sections appear in the Discovery & Evaluation Behavior section.
3. Identify 2–5 ICP candidates from research evidence — note who they are, pain evidence, accessibility, value potential, and WTP signal strength.
4. Checkpoint 1 — Present candidates to user. Show ICP candidates with rationale — cite pain evidence found, accessibility signals, and value delivery reasoning from search findings for each candidate. Ask: "Do any surprise you? Any segment I'm missing?" Incorporate feedback before proceeding.
5. Deep research per ICP: Targeted WebSearch to fill the 9-section framework for each candidate:
   • Customer Profile — buyer persona, budget, discovery channels. Include Geographic Focus (only if the product has regulatory, language, or market-specific constraints — initial target region, why, expansion sequence), Named Accounts (B2B: 5–10 real companies that fit, with name, size, industry, and why they fit), and Business Model & Go-to-Market Motion (model type with evidence, primary motion PLG/SLG/community/partner/hybrid, buyer-user relationship)
   • User Profile(s) — daily users, sophistication, goals, frustrations
   • Trigger Events — what causes them to look NOW? Job changes, growth milestones, compliance deadlines, tool sunsets, contract renewals, funding events, new regulations. Rank by frequency and urgency.
   • Current State Journey — step-by-step current workflow
   • Pain Map — breakdowns, severity, frequency
   • Current Alternatives (User Perspective) — what users say they currently use or have tried, in their own words
   • Market Sizing — TAM/SAM/SOM via bottom-up (company counts × deal size) and top-down (industry reports, competitor revenue). Flag confidence level.
   • Stated Value Drivers — what customers say matters in their own language, the "aha moment" as they describe it. Include Willingness-to-Pay Signals as a bounded evidence subsection: paid alternatives, budget owner/context, current spend or time-cost proxy, switching-cost tolerance, economic urgency, and pricing sensitivity cues. Do not recommend prices, packages, or monetization strategy here.
   • Customer ↔ User Dynamics — post-purchase buyer-user relationship (provisioning, onboarding, admin vs end-user dynamics). For B2B, detailed buying process and DMU live in Discovery & Evaluation Behavior; this section focuses on the post-purchase relationship.
   • Discovery & Evaluation Behavior — how this persona found, evaluated, and chose solutions. Capture behavioural signals only (where they searched, who they asked, what they compared). Populate section 10.
6. Checkpoint 2 — Present scoring matrix and primary ICP selection. Show Value x Accessibility scores. The value rationale must explicitly distinguish pain intensity from WTP quality: active spend on alternatives, clear budget owner/context, high cost of inaction, tolerance for switching costs, and urgency tied to measurable economic outcomes count as strong WTP evidence; verbal interest without budget, free-only behavior, unclear owner, or price sensitivity that outweighs pain count as weak WTP evidence. Ask: "Which constraints, missing segments, or weak evidence should change this ranking?" Incorporate feedback.
7. Cross-ICP analysis: Shared pains, conflicts, product line recs, build sequence, lowest-hanging-fruit x most-value prioritization, discovery & evaluation comparison (how discovery and evaluation behavior differs across ICPs; do different ICPs find and choose solutions through different paths?).
   • Convert secondary ICPs, product-line recommendations, and materially different Cross-ICP Analysis outcomes into research/.progress.yaml product_paths[] entries when they imply a different product surface, product-path scope, audience-first path, or future pivot.
   • Manifest entries must include id, label, source_skill: icp, scope_path, status, reason, archive_reason, archived_at, promoted_at, evidence_refs, revisit_trigger, next_skill, last_touched, and pipeline_stage: icp.
   • Keep the selected primary ICP in active_paths by default. Recommend $product-line activate when secondary ICPs suggest a materially different product surface worth exploring again. Mark non-selected ICP/product paths status: deferred or status: revisit_candidate; do not run full competitive analysis, positioning, journey mapping, UX, or specs for every deferred path unless the user activates one.
8. Checkpoint 3 — Present cross-ICP analysis and build sequence. Show shared pains with source data, conflicts with specific examples, and build sequence rationale grounded in the scoring matrix. Ask: "Does this sequencing make sense?" Incorporate feedback.
9. Final review: Present complete findings summary. Ask: "Ready to write? Anything to adjust?" Only write after user confirms.

After writing is complete, repeat the Recommended next step from the generated ## Next Steps section in the final chat response.

Deliverables

• research/icp.md — Primary ICP in canonical 9 top-level ## sections (Customer Profile, User Profile(s), Trigger Events, Current State Journey, Pain Map, Current Alternatives (User Perspective), Market Sizing, Stated Value Drivers, Customer ↔ User Dynamics). Include WTP evidence as a subsection inside ## Stated Value Drivers, not as a new top-level parser-breaking section. Then supplementary ## Discovery & Evaluation Behavior (section 10: how they find solutions, how they evaluate, how they choose), then ## Additional ICPs (condensed 9-section + condensed discovery & evaluation per ICP, including condensed WTP evidence), then ## Cross-ICP Analysis (prioritization matrix, shared pains, conflicts, product line recs, build sequence, discovery & evaluation comparison), then ## Signals for Downstream Research (unvalidated observations routed to /competitive-analysis, /positioning, /monetization, /gtm)
• research/icp-search-log.md — Raw research log: every query, findings, evidence, scoring rationale, and Marketplace Side Preflight validation/refutation plus side coverage/exclusion rationale
• research/.progress.yaml — product-path manifest when secondary ICPs, Cross-ICP Analysis, or product-line recommendations create parked or promotable paths. Status values include active, deferred, archived, promoted, and revisit_candidate.
• Product-path or cross-path mode: research/{slug}/icp.md + research/{slug}/icp-search-log.md per product path, plus unified research/icp.md cross-referencing all product-path ICPs with top-level prioritization when multiple active paths exist

The output file must end with a ## Next Steps section based on which files exist. Include a Recommended item (the single highest-impact next step given current project state) with a one-line reason, followed by Other options (2–4 alternatives). Use this format in the output:

Next Steps

Recommended: $competitive-analysis — maps the landscape your ICP operates in so positioning and GTM have competitive grounding

Other options:

• $competitive-analysis — Map the competitive landscape for this ICP's market (if no research/competitive-analysis.md)
• IF no research/journey-map.md: check .agents/project.json.enabled_packs for customer-lifecycle — if customer-lifecycle is not enabled, recommend $pack install customer-lifecycle first; if customer-lifecycle is enabled, recommend $journey-map — map the current-state journey to find intervention points
• IF codebase exists: check .agents/project.json.enabled_packs for business-ops — if business-ops is not enabled, recommend $pack install business-ops first; if business-ops is enabled, recommend $mvp-gap — check if the codebase delivers on the ICP's top pain point
• IF research/competitive-analysis.md exists: check .agents/project.json.enabled_packs for product-design — if product-design is not enabled, recommend $pack install product-design first; if product-design is enabled, recommend $brainstorm — generate solution ideas grounded in the ICP research

Only include items whose conditions are met. The recommendation ($competitive-analysis) is always applicable.

Task Classification

When this skill produces follow-up work, file it by execution semantics:

• Immediately actionable implementation or documentation work goes in tasks/todo.md.
• Human-only external actions tied to automated steps go in tasks/manual-todo.md with _(blocks: Step N.X)_ or _(after: Step N.X)_; repo edits, SDK wiring, generated assets, local commands, tests, audits, and authenticated CLI/API work stay in tasks/todo.md.
• One-time condition-gated records, baselines, or future measurements go in tasks/record-todo.md with source, condition, non-blocking reason, evidence, and promotion rule.
• Cadence-based reviews, playtests, adoption checks, investor updates, retros, or docs-health checks go in tasks/recurring-todo.md with cadence, owner/agent, next due, evidence path, and escalation conditions.
• Do not put non-blocking records or recurring obligations in tasks/todo.md unless they have been explicitly promoted into current execution work.

Constraints

• Stay in problem space — do not propose features, architecture, or solutions.
• Evidence-based — every claim must trace to research logged in research/icp-search-log.md.
• Primary ICP must use canonical 9 ## headers for downstream compatibility ($spec-interview, $mvp-gap, $roadmap, $journey-map). The renamed sections (Current Alternatives (User Perspective), Stated Value Drivers) replace the former Market Landscape and Value Proposition headers. Section 10 (## Discovery & Evaluation Behavior) is supplementary and does not affect downstream parsing.
• Section 10 captures behavioural signals only — how personas find, evaluate, and choose solutions. Do not include GTM strategy, channel analysis, budget authority, procurement process, or pricing expectations — those belong in downstream skills ($gtm, $monetization, $enterprise-icp).
• WTP belongs in ICP only as evidence of segment fit and urgency: budget ownership, paid alternatives, current spend/time-cost proxies, switching-cost tolerance, economic urgency, and price sensitivity cues. Do not convert WTP evidence into pricing recommendations, packaging, discounting, ARPA targets, or monetization strategy; route those raw signals to $monetization.
• Minimum 8 WebSearch queries before identifying candidates, 2–3 per candidate after.
• Do not overwrite existing research/icp.md without asking.
• Present before writing — never write output until findings are validated through checkpoints.
• ## Next Steps must be the final section in the output file, with a recommended next step and 2–4 other options.

Alignment Page

When this skill produces durable deliverables (research, specs, plans, reports, prototypes, or any document output), build a full-depth HTML alignment page following ALIGNMENT-PAGE.md in this skill's directory. Output: alignment/icp-{topic}.html.

Default Shipping Contract

Follow the shared shipping contract convention in CLAUDE.md.