Justice — Product Demo: Context Memory
Purpose
A demo that reuses facts the user shared earlier is dramatically more persuasive than one using generic examples. This skill defines how the assistant captures, stores, and applies session context during a product demo, creating a coherent "this was built for me" experience.
Inputs / signals to capture
Listen throughout the demo conversation for the following and store them in the session context object:
| Signal |
Example |
Store as |
| Industry / sector |
"we do MSAs for marketing agencies" |
industry: marketing, doc_type: MSA |
| Firm type |
"I'm a solo practitioner", "we're a 20-person firm" |
firm_size: solo or firm_size: 20 |
| Jurisdiction |
"most of our clients are in Lebanon" |
jurisdiction: LB |
| Practice area |
"we focus on corporate M&A" |
practice_area: corporate_ma |
| Pain point |
"contract review takes us forever" |
pain_point: contract_review_time |
| Current tools |
"we use Word and email" |
current_stack: word, email |
| Language preference |
responds in Arabic |
language: ar |
| Specific documents |
"we do employment contracts a lot" |
doc_focus: employment_contracts |
| Deal size or volume |
"about 30 contracts a month" |
volume: 30_monthly |
Memory rules
Capture
- Capture context passively — do not run a questionnaire at the start. Let context emerge naturally.
- If the user explicitly says something ("I work in construction law in Saudi Arabia"), treat it as high-confidence context.
- If context is implied ("our Riyadh clients" → likely KSA jurisdiction), treat it as lower-confidence; confirm if needed.
Apply
- In all subsequent demo examples, substitute the captured context for generic placeholders:
- Generic: "Let me show you an NDA draft"
- Personalized: "Let me show you an employment contract draft — typical for a Saudi-based firm like yours"
- When invoking a skill for demonstration, pre-fill jurisdiction, document type, and language from context
- When suggesting next steps, reference the user's stated pain points ("Given that contract review is your main bottleneck, here's what the review workflow looks like")
Do not re-ask
- If the user already stated their jurisdiction, do not ask again later in the session
- If the user said they are a solo practitioner, do not show team-workflow steps as the primary demo
Memory scope
- Session only: context memory lives for the duration of the demo conversation
- Cleared at session close unless the user creates an account — do not persist to local storage or backend without user consent
- If the user signs up, prompt: "Shall I save your preferences to your new account?" (opt-in)
Integration with demo flow
The context memory feeds into:
- Example generation: all example contracts, clauses, and research topics are drawn from the user's stated industry and jurisdiction
- Skill suggestions (Step 4 of the demo flow): skills shown are filtered to the user's practice area
- ROI framing: time-saved estimates reference the user's stated volume ("at 30 contracts/month, this saves you ~40 hours")
- Demo recap: [[justice-product-demo-recap]] uses the session context to generate a tailored wrap-up
Failure modes
| Situation |
Handling |
| Context is ambiguous or contradictory |
Use the more recent / more specific signal; proceed with a soft assumption and confirm if high-stakes |
| No context captured by Step 3 of demo |
Ask one targeted question: "What type of work do you do most — drafting, reviewing, or research?" |
| User changes context mid-demo |
Update immediately; acknowledge the change naturally ("Got it — let me switch to a construction contract example instead") |
Related skills
- [[justice-product-demo-recap]]
- [[justice-intent-product-demo-request]]
- [[justice-result-page-routing]]
1---2name: justice-product-demo-context-memory3description: Use during a product demo session to retain and reuse context the user shared earlier in the session — industry, firm type, key facts, stated preferences — so that subsequent demo examples feel tailored rather than generic. Session memory is ephemeral (ends at session close unless the user signs up). Works alongside the demo recap and deep-link skills to create a coherent, personalized demo experience.4license: MIT5metadata:6 id: justice.product-demo.context-memory7 category: justice8 jurisdictions: [__multi__]9 priority: P210 intent: [demo, memory, personalization, session-context]11 related: [justice-product-demo-recap, justice-intent-product-demo-request, justice-result-page-routing]12 source: https://skillmd.com/skills/sboghossian-mini-claude-for-legal/justice-product-demo-context-memory13 upstream: Louis — HAQQ Legal AI (github.com/sboghossian/mini-claude-for-legal)14 version: "1.0"15---1617# Justice — Product Demo: Context Memory1819## Purpose2021A demo that reuses facts the user shared earlier is dramatically more persuasive than one using generic examples. This skill defines how the assistant captures, stores, and applies session context during a product demo, creating a coherent "this was built for me" experience.2223## Inputs / signals to capture2425Listen throughout the demo conversation for the following and store them in the session context object:2627| Signal | Example | Store as |28|---|---|---|29| **Industry / sector** | "we do MSAs for marketing agencies" | `industry: marketing`, `doc_type: MSA` |30| **Firm type** | "I'm a solo practitioner", "we're a 20-person firm" | `firm_size: solo` or `firm_size: 20` |31| **Jurisdiction** | "most of our clients are in Lebanon" | `jurisdiction: LB` |32| **Practice area** | "we focus on corporate M&A" | `practice_area: corporate_ma` |33| **Pain point** | "contract review takes us forever" | `pain_point: contract_review_time` |34| **Current tools** | "we use Word and email" | `current_stack: word, email` |35| **Language preference** | responds in Arabic | `language: ar` |36| **Specific documents** | "we do employment contracts a lot" | `doc_focus: employment_contracts` |37| **Deal size or volume** | "about 30 contracts a month" | `volume: 30_monthly` |3839## Memory rules4041### Capture42- Capture context passively — do not run a questionnaire at the start. Let context emerge naturally.43- If the user explicitly says something ("I work in construction law in Saudi Arabia"), treat it as high-confidence context.44- If context is implied ("our Riyadh clients" → likely KSA jurisdiction), treat it as lower-confidence; confirm if needed.4546### Apply47- In all subsequent demo examples, substitute the captured context for generic placeholders:48 - Generic: "Let me show you an NDA draft"49 - Personalized: "Let me show you an employment contract draft — typical for a Saudi-based firm like yours"50- When invoking a skill for demonstration, pre-fill jurisdiction, document type, and language from context51- When suggesting next steps, reference the user's stated pain points ("Given that contract review is your main bottleneck, here's what the review workflow looks like")5253### Do not re-ask54- If the user already stated their jurisdiction, do not ask again later in the session55- If the user said they are a solo practitioner, do not show team-workflow steps as the primary demo5657### Memory scope58- **Session only**: context memory lives for the duration of the demo conversation59- **Cleared at session close** unless the user creates an account — do not persist to local storage or backend without user consent60- If the user signs up, prompt: "Shall I save your preferences to your new account?" (opt-in)6162## Integration with demo flow6364The context memory feeds into:65661. **Example generation**: all example contracts, clauses, and research topics are drawn from the user's stated industry and jurisdiction672. **Skill suggestions** (Step 4 of the demo flow): skills shown are filtered to the user's practice area683. **ROI framing**: time-saved estimates reference the user's stated volume ("at 30 contracts/month, this saves you ~40 hours")694. **Demo recap**: [[justice-product-demo-recap]] uses the session context to generate a tailored wrap-up7071## Failure modes7273| Situation | Handling |74|---|---|75| Context is ambiguous or contradictory | Use the more recent / more specific signal; proceed with a soft assumption and confirm if high-stakes |76| No context captured by Step 3 of demo | Ask one targeted question: "What type of work do you do most — drafting, reviewing, or research?" |77| User changes context mid-demo | Update immediately; acknowledge the change naturally ("Got it — let me switch to a construction contract example instead") |7879## Related skills8081- [[justice-product-demo-recap]]82- [[justice-intent-product-demo-request]]83- [[justice-result-page-routing]]