bizanalysis

name	bizanalysis
description	Analyzes and critiques business viability. Use after /brainstorm and before /prd.
argument-hint	[output file path, default docs/business_analysis.md]
disable-model-invocation	false
allowed-tools	Task, Read, Glob, Grep, Write, Edit, WebSearch, Bash(python3 scripts/has_skill.py *), Bash(python3 scripts/capture_source.py *), Bash(python3 scripts/validate_research_claim.py *), Bash(python3 scripts/synthesize_from_deep_research.py *), Bash(bash ${CLAUDE_PLUGIN_ROOT}/scripts/*), Bash(python3 ${CLAUDE_PLUGIN_ROOT}/scripts/*)

Kit Preamble — bizanalysis

Kit Script Root

Kit root: ${CLAUDE_PLUGIN_ROOT}

• Absolute path above → plugin install (substituted at load time; no project scripts/ dir): prefix every kit script command with it, e.g. bash <kit-root>/scripts/checkpoint.sh …. Absolute paths also work from worktrees.
• Literal ${…} placeholder above → standalone layout: run commands as written.

Project Context Detection

Run these checks silently at the start. Use results to adapt behavior:

• [ -f issues.md ] — if true, this project uses the sprint system. Respect issue numbering and STATUS.md.
• [ -f docs/sprint_state.md ] — if true and Status shows running, a sprint is active. Be aware of parallel work in worktrees.
• [ -f docs/prd_digest.md ] — if true, read it for quick project context before starting.

Kit Rules

• Verify gh auth status before any GitHub operation. Steps:

1. Determine the output path ($ARGUMENTS or docs/business_analysis.md).
2. Check if the file already exists at that path.
3. Check if docs/brainstorm_notes.md exists. If it does, read it and use as context — problem space, target users, chosen direction.
4. Probe runtime (per SPEC-018 platform-first delegation): python3 scripts/has_skill.py deep-research. Branch:
   • Exit 0 OR 2 → primary path (steps 5–9 use /deep-research per dimension).
   • Exit 1 → degraded path (steps 5–9 use capture_source.py + validate_research_claim.py per source).

If the file does NOT exist (New Analysis):

5a) If brainstorm notes exist, summarize the idea from them. Otherwise, ask the user to describe their idea.

Primary path (runtime probe = 0 or 2): 6a-P) Market Analysis: invoke /deep-research with a refined question about market size, trends, and TAM/SAM/SOM for the specific segment. Emit telemetry event research_delegated_to_deep_research. 7a-P) Competitive Landscape: invoke /deep-research with a refined question about direct competitors, alternatives, and positioning differences. Emit research_delegated_to_deep_research. 8a-P) Business Model + Risks: invoke /deep-research with a refined question about prevailing revenue models in the segment and the key market / regulatory / competitive risks. Emit research_delegated_to_deep_research. 9a-P) Synthesize: pass the per-dimension /deep-research reports through python3 scripts/synthesize_from_deep_research.py --mode bizanalysis. The synthesizer maps cited claims into the kit's 5-section template, preserves [single-source] / [contested] flags verbatim from upstream, and renders empty sections as the literal line Data: not available — re-run /deep-research with a sharper question or accept "no data". 10a-P) Audit: invoke Task with subagent_type: synthesizer-auditor, inputs = (rendered draft, all /deep-research reports). Block save on claim_dropped / claim_distorted / evidence_distorted / scope_change findings until resolved.

Degraded path (runtime probe = 1): 6a-D) Market Analysis: ask the user for source URLs about market size + trends. For each: python3 scripts/capture_source.py <url>. Construct each claim per templates/research_claim.md. TAM/SAM/SOM values with only 1 distinct-domain source render as range: <low–high> [single-source], not a point number. Emit research_degraded_path_used once at the start of the path. 7a-D) Competitive Landscape: capture sources for each competitor mentioned. Build a comparison table where every cell with a quantitative claim carries a claim record. 8a-D) Business Model + Risks: capture sources or render the no-data literal where no source is available. 9a-D) Validate: for every claim authored on this path, run python3 scripts/validate_research_claim.py --claim-file <claim>.json. Block save on any non-ok verdict (quote_missing / stale / snapshot_absent / bad_input). 10a-D) Audit: invoke Task with subagent_type: research-auditor, inputs = (rendered draft, docs/references/research/). Block save on context_contradicts / scope_change findings.

11a) Critique: Produce a SWOT analysis and a Go/Pivot/No-Go recommendation. The SWOT is authored synthesis over the validated claims; it inherits the no-data literal for any dimension where the underlying analysis section is empty. Discuss the recommendation with the user.

If the file DOES exist (Continue Analysis):

5b) Read the existing analysis and present a brief recap. 6b) Ask the user which section they want to develop, challenge, or update. 7b) Repeat the chosen dimension via the same primary/degraded path the runtime probe selects. Never re-introduce free-form web-page claim extraction on either path — both delegation and capture-snapshot flows remain mandatory (SPEC-018). 8b) Merge new insights into the existing analysis. The audit step (synthesizer-auditor or research-auditor) re-runs after the merge.

Common (both modes):

12. Present the draft to the user and incorporate feedback.
13. Save the final analysis to the output path with the five sections (Executive Summary / Market Analysis / Competitive Landscape / Business Model / Risks & Mitigations). The synthesizer enforces canonical section order regardless of input order.
14. Inform the user they can run /prd next — business analysis and brainstorm notes will be automatically used as context.

Error Handling

• If the output path is not writable: report the error and ask for an alternative path.
• If has_skill.py reports exit code 2 (runtime-built-in, filesystem-invisible) and /deep-research returns a "skill not found" error inline, fall back to the degraded path and emit research_degraded_path_used.
• If the synthesizer-auditor or research-auditor returns an audit_confidence: low, surface its summary to the user before save and confirm intent rather than silently proceeding.

Quality Criteria

NEVER:

• Fabricate market size numbers. Every quantitative claim must carry a Source: line that traces back to either a /deep-research report (primary path) or a captured snapshot under docs/references/research/ (degraded path).
• Unconditionally praise the user's idea — provide honest critique.
• Ask more than 3 questions at a time — keep it conversational.
• Jump to conclusions without analysis — show your reasoning.
• Skip the audit step. The validator catches quote presence; only the auditor catches context-mismatched usage.

INSTEAD:

• Cite sources verbatim through the synthesizer (primary) or the claim record (degraded).
• Present both strengths and weaknesses.
• Use concrete numbers AND their source citation — never numbers in isolation.
• Render the no-data literal explicitly when sources are unavailable.
• Propose actionable next steps for each risk identified.

Limits

/deep-research's multi-agent adversarial verification (primary path) lowers the research-correctness floor far below anything the kit could build inline. The synthesizer-auditor catches kit-side paraphrase distortion. Together they push fabrication to near-zero, but the floor is not zero: a claim that is verbatim in the upstream report AND not directly contradicted by surrounding context can still be misinterpreted by the user (conditional forecast cited as fact, scope mismatch). Document this honestly in the Risks section of the output; do NOT promise zero fabrication in any user-facing message.

Guidelines

• This is an interactive, conversational skill — engage naturally.
• If the user says "that's enough" or similar, produce the best analysis possible with only the validated claims. If a section's sources are unavailable, render the no-data literal rather than ship paraphrased claims.
• After saving, suggest next step: /prd to create a PRD using both business analysis and brainstorm notes as context.