code-ai-installer 3.0.0 → 4.0.0

package/domains/analytics/locales/en/.agents/skills/session-prompt-generator/SKILL.md

@@ -1,378 +1,378 @@
----
-name: session-prompt-generator
-description: Generation of the prompt for the next session — context gathering, prompt formulation, progress saving
----
-# Session Prompt Generator — Generating a Prompt for the Next Session
-
-## When to Use
-- When **completing a session** in the Full Pipeline (`/analyze`) — before the inter-session `$handoff`.
-- When **needing to hand off the project** to another operator — forming a self-sufficient prompt.
-- Upon **user request** "generate a prompt to continue" — at any point.
-- During an **abnormal termination** of a session — saving maximum context for recovery.
-
-> **Key Principle:** The next session starts with **ZERO** context. The prompt must contain everything necessary for the new session to continue work without losses.
-
-## Input
-
-| Field | Required | Description |
-|------|:-----------:|----------|
-| `$board` state | ✅ | Current task board with all statuses |
-| Completed deliverables | ✅ | Full artifacts of the current and previous sessions |
-| Current gate | ✅ | Where we stopped (gate ID from `$board`) |
-| Next gate | ✅ | Where we are moving (next gate ID) |
-| Open questions | ⬚ | Blockers, unresolved conflicts, questions for the user |
-| Current session number | ✅ | Which session is ending (1-5) |
-| Project name | ✅ | Name of the analytical project |
-
-### Relationship with Other Skills
-| Skill | Interaction | When |
-|------|----------------|-------|
-| `$board` | Source: the current board state is included in the prompt (block 5) | Step 0 |
-| `$handoff` | The prompt is included in the inter-session handoff file | Step 5 |
-| `$gates` | Source: requirements for the next gate define the task (block 6) | Step 0 |
-| Framework skills | Completed frameworks are included as full text in block 4 | Step 1 |
-
-## Mapping "Session → Gates → Agents → Skills"
-
-### Full Pipeline (`/analyze`)
-
-| Session | Gates | Agents | Required Skills | Framework Skills |
-|:------:|-------|--------|-------------------|-----------------|
-| **1** | COND-01, INT-01 | Conductor, Interviewer | `$board`, `$gates`, `$handoff` | — |
-| **2** | COND-02, RES-01, AN-01, ST-01 | Conductor, Researcher, Data Analyst, Strategist | `$board`, `$gates`, `$handoff`, `$web-research` | By AN-01 choice (min 2) |
-| **3** | COND-03, RES-02, AN-02, ST-02 | Conductor, Researcher, Data Analyst, Strategist | `$board`, `$gates`, `$handoff`, `$web-research` | By AN-02 choice (counter-frameworks) |
-| **4** | COND-04, MED-01 | Conductor, Mediator | `$board`, `$gates`, `$handoff` | — |
-| **5** | COND-05, DS-01, LY-01, RG-01 | Conductor, Designer, Layouter, Conductor (RG) | `$board`, `$gates`, `$handoff`, `$report-design`, `$html-pdf-report` | — |
-
-### Quick Pipeline (`/quick-insight`)
-One session — no prompt needed (all gates are within a single session).
-
-## Protocol
-
-### Step 0 — Gathering State
-1. Read the current `$board` state — all gates, statuses, deliverables.
-2. Collect all completed artifacts of the current session.
-3. Load the cumulative index from previous sessions (if any).
-4. Determine the next gate and its requirements from `$gates`.
-5. Use the "Session → Gates → Agents → Skills" mapping to determine what to output.
-
-### Step 1 — Identifying Critical Data
-1. Highlight data that is **irrecoverable** without previous context:
-
-   | Category | Examples | Priority |
-   |-----------|---------|:---------:|
-   | User responses | Interview (INT-01), clarifications, decisions | 🔴 Critical |
-   | Collected data | RES-xx data with URLs and sources | 🔴 Critical |
-   | Analysis results | AN-xx frameworks, scoring, findings | 🔴 Critical |
-   | Strategic recommendations | ST-xx strategy, action plan | 🔴 Critical |
-   | Mediator's verdict | MED-01 conclusion, reasoning | 🔴 Critical |
-   | User decisions | Approved, rejections, scope changes | 🟡 Important |
-   | Progress metrics | Time per gate, blockers | 🟢 Nice-to-have |
-
-2. **Rule:** All 🔴 Critical data is included **as full text**. No exceptions.
-
-### Step 2 — Determining Agents and Skills
-1. Use the mapping to determine: which agents and skills are needed in the next session.
-2. Form a list of files to load (specific paths).
-3. For framework skills: specify concrete skills if already known (from AN-01 or research brief).
-
-### Step 3 — Forming the Prompt
-1. Fill out the prompt template (6 blocks — see below).
-2. Include the FULL text of critical data (no links, no summaries).
-3. Formulate a specific task for the next session.
-4. Specify which agents and skills to activate (specific file paths).
-
-### Step 4 — Prompt Validation
-Self-sufficiency test — for each point, answer "is it possible without external context?":
-
-| # | Check | Result |
-|---|----------|:---------:|
-| 1 | Can work start with ONLY this prompt? | Yes / No |
-| 2 | Are there no references to "previous context" / "as discussed"? | Yes / No |
-| 3 | Is all 🔴 Critical data included as full text? | Yes / No |
-| 4 | Is the task for the session specific (gate, action, deliverable)? | Yes / No |
-| 5 | Are agents and skills listed with file paths? | Yes / No |
-| 6 | Is the task board included? | Yes / No |
-| 7 | Are open questions noted? | Yes / No |
-
-**If at least one = No — refine the prompt.**
-
-### Step 5 — Saving
-1. Include the prompt in the `$handoff` file: `docs/analytics/session-N-handoff.md`.
-2. Present the prompt to the user to save / copy.
-3. Warn: "This is the prompt — insert it at the beginning of the next session."
-
-### Step 6 — Emergency Mode
-If the session ends abruptly (timeout, error, context loss):
-
-1. **Immediately** save everything available: the board, the latest deliverable, open questions.
-2. Form an **emergency prompt** (minimal version):
-   - Block 1 (identification) — mandatory.
-   - Block 5 (board state) — mandatory.
-   - Block 4 (data) — everything that could be collected.
-   - Block 6 (task) — "Restore context from session-N-handoff.md, check data integrity, continue from gate [ID]."
-3. Mark the prompt as **⚠️ EMERGENCY — possible data loss**.
-4. Indicate what data might have been lost and how to recover it (repeated `$web-research`, asking the user).
-
-## Generated Prompt Structure
-
-### Block 1 — Project Identification
-- Project name.
-- Pipeline mode (Full / Quick).
-- Next session number.
-- Which sessions are completed.
-
-### Block 2 — Loading Instruction
-- Read AGENTS.md of the analytics domain.
-- Load agents needed in this session (from mapping).
-- Load skills: `$board`, `$handoff`, `$gates` + gate-specific.
-- Specific file paths.
-
-### Block 3 — Completed Work (Summary)
-- What was done in previous sessions (2-3 sentences per session).
-- Key decisions and their reasoning.
-
-### Block 4 — Critical Data
-- FULL TEXT of key artifacts needed to continue.
-- Research data, analysis findings, recommendations.
-- User responses to the interview (if any).
-
-### Block 5 — Current State
-- Task board (`$board`) in its current state.
-- What the next gate is.
-- Open questions and blockers.
-
-### Block 6 — Task for the Session
-- Specific formulation: what needs to be done in this session.
-- Expected result (deliverables).
-- Constraints and specifics.
-
-## Prompt Generation Rules
-
-### Mandatory Rules
-1. **Self-sufficiency** — the prompt must work without ANY previous context.
-2. **Data completeness** — include FULL text, not links and not summaries.
-3. **Task specificity** — not "continue analysis", but "execute RES-02: research [topic] from the Critics' perspective, Analysts' data: [data]".
-4. **Structure** — clear blocks, marked with headings.
-5. **Skill loading** — explicitly state which agents and skills to load with file paths.
-
-### Forbidden Elements
-1. Links to "previous session" without full content.
-2. "As we discussed" — there is no context, nothing to discuss.
-3. "See results above" — there is no "above", everything must be in the prompt.
-4. Abbreviated data — "main findings: ...". ALL findings are needed.
-5. Assumptions about context — "you already know that...".
-
-## Example — Prompt for Session 3 (Team Beta: Critics)
-
-```markdown
----
-## Prompt for Session 3 — Analysis of EdTech Market RU
-
-### 1. Project
-- **Name:** Analysis of EdTech Market RU for corporate training
-- **Mode:** Full Pipeline (/analyze)
-- **Session:** 3 of 5
-- **Previous Sessions:** 1 (Interview + Brief), 2 (Team Alpha: Research + Analysis + Strategy)
-
-### 2. Loading
-Read:
-- `domains/analytics/AGENTS.md`
-- `domains/analytics/.agents/skills/board/SKILL.md`
-- `domains/analytics/.agents/skills/handoff/SKILL.md`
-- `domains/analytics/.agents/skills/gates/SKILL.md`
-- `domains/analytics/.agents/skills/web-research/SKILL.md`
-- `domains/analytics/.agents/skills/competitive-analysis/SKILL.md`
-- `domains/analytics/.agents/skills/porters-five-forces/SKILL.md`
-
-Activate agents: Conductor (COND-03), Researcher (RES-02), Data Analyst (AN-02), Strategist (ST-02).
-Role: **Team Beta (Critics)** — task: critique and alternative analysis of Team Alpha's results.
-
-### 3. What is Completed
-**Session 1:** Conducted interview, formed research brief (12 questions), defined scope: corporate training, B2B SaaS, RU, 3-year horizon. Evaluation mode: scoring.
-**Session 2:** Team Alpha (Analysts) conducted research (28 sources), applied TAM/SAM/SOM + Competitive Analysis + PEST. Strategic recommendation: "AI-first corporate training platform, focus on 50-500 headcount companies."
-
-### 4. Data (full text)
-
-#### Research Brief (from INT-01)
-1. What is the size of EdTech Market RU (TAM/SAM/SOM) for corporate training?
-2. Who are the key competitors and what are their shares?
-3. What technological trends are changing the market?
-4. What regulatory factors affect EdTech?
-[... all 12 questions ...]
-
-#### Team Alpha Research Data (from RES-01)
-**TAM:** $4.2B (EdTech RU, 2026). Source: Smart Ranking, Jan 2026 ✅
-**SAM:** $1.8B (corporate training). Source: HSE, Mar 2026 ✅
-**SOM:** $180M (B2B SaaS platforms). Source: calculation ⚠️ (2 sources instead of 3)
-
-**Competitors:**
-| Competitor | Share | Threat Score | Key Strength |
-|-----------|------|-------------|---------------|
-| Skillbox (B2B) | 15% | 7.2 | Brand, content |
-| Netology | 12% | 6.8 | Corp. programs |
-| Yandex Praktikum | 18% | 7.6 | Yandex Ecosystem |
-[... full data ...]
-
-#### Team Alpha Analysis Results (from AN-01, ST-01)
-[FULL TEXT: TAM/SAM/SOM framework, Competitive Analysis, PEST, strategic recommendations]
-
-### 5. Current State
-
-#### Task Board
-📊 Progress: 40% (6/15) | Session: 3/5 | Blockers: 0
-
-| # | ID | Gate | Session | Status | Deliverable |
-|---|----|------|:------:|:------:|-------------|
-| 1 | COND-01 | Conductor | 1 | [✓] | Full Pipeline |
-| 2 | INT-01 | Interviewer | 1 | [✓] | Brief: 12 questions |
-| 3 | COND-02 | Conductor (s.2) | 2 | [✓] | Context loaded |
-| 4 | RES-01 | Researcher (Alpha) | 2 | [✓] | 28 sources |
-| 5 | AN-01 | Data Analyst (Alpha) | 2 | [✓] | 3 frameworks |
-| 6 | ST-01 | Strategist (Alpha) | 2 | [✓] | AI-first strategy |
-| 7 | COND-03 | Conductor (s.3) | 3 | [→] | — |
-| 8 | RES-02 | Researcher (Beta) | 3 | [ ] | — |
-| 9 | AN-02 | Data Analyst (Beta) | 3 | [ ] | — |
-| 10 | ST-02 | Strategist (Beta) | 3 | [ ] | — |
-[... remaining gates ...]
-
-#### Open Questions
-- ⚠️ SOM is calculated based on 2 sources (Alpha recommended verifying)
-
-### 6. Task for Session 3
-**Gates:** COND-03 → RES-02 → AN-02 → ST-02
-**What to do:** Team Beta (Critics) must:
-1. **RES-02:** Investigate the EdTech market from alternative perspectives. Check Alpha's data. Find counter-arguments. Verify SOM (find 3rd source).
-2. **AN-02:** Apply alternative frameworks (Porter's 5 Forces, Blue Ocean). Find weaknesses in Alpha's analysis.
-3. **ST-02:** Formulate an alternative strategy. Critique Alpha's strategy with reasoning.
-**Expected Result:** Alternative package (data + analysis + strategy) for handoff to Mediator in session 4.
-
-→ Start as Conductor (COND-03). Load the board, check Alpha's data, pass control to Researcher (RES-02).
----
-```
-
-## Validation (Quality Gate)
-
-A prompt is considered correct if:
-
-- [ ] The prompt is self-sufficient — work can start without external context
-- [ ] Project name and mode are specified
-- [ ] Session number and session history are specified
-- [ ] Agents and skills to load are listed with file paths (from mapping)
-- [ ] All 🔴 Critical data is included as full text
-- [ ] Task board (`$board`) is included in its current state with metrics
-- [ ] The next gate and task are formulated specifically (gate, action, deliverable)
-- [ ] Open questions and blockers are noted
-- [ ] No forbidden elements (links without content, abbreviations, assumptions)
-- [ ] Validation test (7 points, step 4) — all "Yes"
-- [ ] In emergency mode — prompt is marked ⚠️, lost data is indicated
-
-## Handoff
-The output of `$session-prompt-generator` is an input for:
-- **Next session** — via user copying the prompt.
-- **`$handoff`** — the prompt is included in the `docs/analytics/session-N-handoff.md` file.
-- **`$board`** — the board state is saved in the prompt.
-
-Handoff format: prompt in markdown, ready to be pasted at the beginning of the new session.
-
-## Anti-patterns
-
-| Error | Why it's bad | How to do it right |
-|--------|-------------|---------------|
-| Prompt without research data | The next session will lose all collected facts | Include FULL text of data (🔴 Critical) |
-| "Continue work" | No context on what to continue | Specific task: gate, action, deliverable |
-| File links without content | Files might be inaccessible | Include content directly in the prompt |
-| Prompt without task board | Unclear where we are in the pipeline | Always include `$board` with metrics |
-| Abbreviated artifacts | Critical details are lost | Full text of every 🔴 Critical artifact |
-| Skipping open questions | Problems will resurface | Note all blockers and questions |
-| No loading instructions | The session will not load the needed agents | Mapping → specific file paths |
-| Prompt > 50% "fluff" | Dilution of critical data | Structure: blocks, tables, specifics |
-| Abnormal termination without prompt | Total loss of context | Emergency mode: minimal prompt + ⚠️ mark |
-
-## Output Template
-
-```markdown
----
-## Prompt for Session [N+1] — [Project Name]
-
-### 1. Project
-- **Name:** [Name]
-- **Mode:** Full Pipeline (/analyze)
-- **Session:** [N+1] of 5
-- **Previous Sessions:** 1-[N] completed
-
-### 2. Loading
-Read:
-- `domains/analytics/AGENTS.md`
-- `domains/analytics/.agents/skills/board/SKILL.md`
-- `domains/analytics/.agents/skills/handoff/SKILL.md`
-- `domains/analytics/.agents/skills/gates/SKILL.md`
-- [Specific skills: file paths]
-
-Activate agents: [List of agents with Gate IDs]
-
-### 3. What is Completed
-**Session 1:** [Summary 2-3 sentences]
-**Session [N]:** [Summary 2-3 sentences]
-
-**Key decisions:**
-| # | Decision | Reasoning | Adopted by |
-|---|---------|-------------|------------|
-| 1 | [Decision] | [Why] | [User / Conductor] |
-
-### 4. Data (full text)
-
-#### Research Brief (from INT-01)
-[FULL TEXT research brief]
-
-#### Research Data (from RES-xx)
-[FULL TEXT data with sources and URLs]
-
-#### Analysis Results (from AN-xx)
-[FULL TEXT frameworks and findings]
-
-#### [Other artifacts as needed]
-[FULL TEXT]
-
-### 5. Current State
-
-#### Task Board
-[Full board from $board with progress metrics]
-
-#### Open Questions
-| # | Question / Blocker | Severity | For which gate |
-|---|-----------------|:--------:|:----------------:|
-| 1 | [Description] | 🔴/🟡/🟢 | [ID] |
-
-### 6. Task for Session [N+1]
-**Gates:** [ID list]
-**What to do:** [Specific description per gate]
-**Expected Result:** [Deliverables]
-**Constraints:** [If any]
-
-→ Start as Conductor (COND-xx). Load the board, check the data, pass control to [next agent].
----
-```
-
-### Emergency Prompt (minimal version)
-
-```markdown
----
-## ⚠️ EMERGENCY PROMPT — Session [N+1] — [Project Name]
-
-**Status:** Session [N] terminated abnormally. Possible data loss.
-
-### What is Saved
-[Everything collected: board, latest deliverable, questions]
-
-### What might be lost
-| # | Data | Recovery Method |
-|---|--------|-----------------------|
-| 1 | [Lost data] | [Repeated web-research / ask user / recover from handoff] |
-
-### Task
-Restore context from `docs/analytics/session-[N]-handoff.md`. Check data integrity. Continue from gate [ID].
-
-→ Start as Conductor (COND-xx). Priority: verify data integrity before continuing.
----
-```
+---
+name: session-prompt-generator
+description: Generation of the prompt for the next session — context gathering, prompt formulation, progress saving
+---
+# Session Prompt Generator — Generating a Prompt for the Next Session
+
+## When to Use
+- When **completing a session** in the Full Pipeline (`/analyze`) — before the inter-session `$handoff`.
+- When **needing to hand off the project** to another operator — forming a self-sufficient prompt.
+- Upon **user request** "generate a prompt to continue" — at any point.
+- During an **abnormal termination** of a session — saving maximum context for recovery.
+
+> **Key Principle:** The next session starts with **ZERO** context. The prompt must contain everything necessary for the new session to continue work without losses.
+
+## Input
+
+| Field | Required | Description |
+|------|:-----------:|----------|
+| `$board` state | ✅ | Current task board with all statuses |
+| Completed deliverables | ✅ | Full artifacts of the current and previous sessions |
+| Current gate | ✅ | Where we stopped (gate ID from `$board`) |
+| Next gate | ✅ | Where we are moving (next gate ID) |
+| Open questions | ⬚ | Blockers, unresolved conflicts, questions for the user |
+| Current session number | ✅ | Which session is ending (1-5) |
+| Project name | ✅ | Name of the analytical project |
+
+### Relationship with Other Skills
+| Skill | Interaction | When |
+|------|----------------|-------|
+| `$board` | Source: the current board state is included in the prompt (block 5) | Step 0 |
+| `$handoff` | The prompt is included in the inter-session handoff file | Step 5 |
+| `$gates` | Source: requirements for the next gate define the task (block 6) | Step 0 |
+| Framework skills | Completed frameworks are included as full text in block 4 | Step 1 |
+
+## Mapping "Session → Gates → Agents → Skills"
+
+### Full Pipeline (`/analyze`)
+
+| Session | Gates | Agents | Required Skills | Framework Skills |
+|:------:|-------|--------|-------------------|-----------------|
+| **1** | COND-01, INT-01 | Conductor, Interviewer | `$board`, `$gates`, `$handoff` | — |
+| **2** | COND-02, RES-01, AN-01, ST-01 | Conductor, Researcher, Data Analyst, Strategist | `$board`, `$gates`, `$handoff`, `$web-research` | By AN-01 choice (min 2) |
+| **3** | COND-03, RES-02, AN-02, ST-02 | Conductor, Researcher, Data Analyst, Strategist | `$board`, `$gates`, `$handoff`, `$web-research` | By AN-02 choice (counter-frameworks) |
+| **4** | COND-04, MED-01 | Conductor, Mediator | `$board`, `$gates`, `$handoff` | — |
+| **5** | COND-05, DS-01, LY-01, RG-01 | Conductor, Designer, Layouter, Conductor (RG) | `$board`, `$gates`, `$handoff`, `$report-design`, `$html-pdf-report` | — |
+
+### Quick Pipeline (`/quick-insight`)
+One session — no prompt needed (all gates are within a single session).
+
+## Protocol
+
+### Step 0 — Gathering State
+1. Read the current `$board` state — all gates, statuses, deliverables.
+2. Collect all completed artifacts of the current session.
+3. Load the cumulative index from previous sessions (if any).
+4. Determine the next gate and its requirements from `$gates`.
+5. Use the "Session → Gates → Agents → Skills" mapping to determine what to output.
+
+### Step 1 — Identifying Critical Data
+1. Highlight data that is **irrecoverable** without previous context:
+
+   | Category | Examples | Priority |
+   |-----------|---------|:---------:|
+   | User responses | Interview (INT-01), clarifications, decisions | 🔴 Critical |
+   | Collected data | RES-xx data with URLs and sources | 🔴 Critical |
+   | Analysis results | AN-xx frameworks, scoring, findings | 🔴 Critical |
+   | Strategic recommendations | ST-xx strategy, action plan | 🔴 Critical |
+   | Mediator's verdict | MED-01 conclusion, reasoning | 🔴 Critical |
+   | User decisions | Approved, rejections, scope changes | 🟡 Important |
+   | Progress metrics | Time per gate, blockers | 🟢 Nice-to-have |
+
+2. **Rule:** All 🔴 Critical data is included **as full text**. No exceptions.
+
+### Step 2 — Determining Agents and Skills
+1. Use the mapping to determine: which agents and skills are needed in the next session.
+2. Form a list of files to load (specific paths).
+3. For framework skills: specify concrete skills if already known (from AN-01 or research brief).
+
+### Step 3 — Forming the Prompt
+1. Fill out the prompt template (6 blocks — see below).
+2. Include the FULL text of critical data (no links, no summaries).
+3. Formulate a specific task for the next session.
+4. Specify which agents and skills to activate (specific file paths).
+
+### Step 4 — Prompt Validation
+Self-sufficiency test — for each point, answer "is it possible without external context?":
+
+| # | Check | Result |
+|---|----------|:---------:|
+| 1 | Can work start with ONLY this prompt? | Yes / No |
+| 2 | Are there no references to "previous context" / "as discussed"? | Yes / No |
+| 3 | Is all 🔴 Critical data included as full text? | Yes / No |
+| 4 | Is the task for the session specific (gate, action, deliverable)? | Yes / No |
+| 5 | Are agents and skills listed with file paths? | Yes / No |
+| 6 | Is the task board included? | Yes / No |
+| 7 | Are open questions noted? | Yes / No |
+
+**If at least one = No — refine the prompt.**
+
+### Step 5 — Saving
+1. Include the prompt in the `$handoff` file: `docs/analytics/session-N-handoff.md`.
+2. Present the prompt to the user to save / copy.
+3. Warn: "This is the prompt — insert it at the beginning of the next session."
+
+### Step 6 — Emergency Mode
+If the session ends abruptly (timeout, error, context loss):
+
+1. **Immediately** save everything available: the board, the latest deliverable, open questions.
+2. Form an **emergency prompt** (minimal version):
+   - Block 1 (identification) — mandatory.
+   - Block 5 (board state) — mandatory.
+   - Block 4 (data) — everything that could be collected.
+   - Block 6 (task) — "Restore context from session-N-handoff.md, check data integrity, continue from gate [ID]."
+3. Mark the prompt as **⚠️ EMERGENCY — possible data loss**.
+4. Indicate what data might have been lost and how to recover it (repeated `$web-research`, asking the user).
+
+## Generated Prompt Structure
+
+### Block 1 — Project Identification
+- Project name.
+- Pipeline mode (Full / Quick).
+- Next session number.
+- Which sessions are completed.
+
+### Block 2 — Loading Instruction
+- Read AGENTS.md of the analytics domain.
+- Load agents needed in this session (from mapping).
+- Load skills: `$board`, `$handoff`, `$gates` + gate-specific.
+- Specific file paths.
+
+### Block 3 — Completed Work (Summary)
+- What was done in previous sessions (2-3 sentences per session).
+- Key decisions and their reasoning.
+
+### Block 4 — Critical Data
+- FULL TEXT of key artifacts needed to continue.
+- Research data, analysis findings, recommendations.
+- User responses to the interview (if any).
+
+### Block 5 — Current State
+- Task board (`$board`) in its current state.
+- What the next gate is.
+- Open questions and blockers.
+
+### Block 6 — Task for the Session
+- Specific formulation: what needs to be done in this session.
+- Expected result (deliverables).
+- Constraints and specifics.
+
+## Prompt Generation Rules
+
+### Mandatory Rules
+1. **Self-sufficiency** — the prompt must work without ANY previous context.
+2. **Data completeness** — include FULL text, not links and not summaries.
+3. **Task specificity** — not "continue analysis", but "execute RES-02: research [topic] from the Critics' perspective, Analysts' data: [data]".
+4. **Structure** — clear blocks, marked with headings.
+5. **Skill loading** — explicitly state which agents and skills to load with file paths.
+
+### Forbidden Elements
+1. Links to "previous session" without full content.
+2. "As we discussed" — there is no context, nothing to discuss.
+3. "See results above" — there is no "above", everything must be in the prompt.
+4. Abbreviated data — "main findings: ...". ALL findings are needed.
+5. Assumptions about context — "you already know that...".
+
+## Example — Prompt for Session 3 (Team Beta: Critics)
+
+```markdown
+---
+## Prompt for Session 3 — Analysis of EdTech Market RU
+
+### 1. Project
+- **Name:** Analysis of EdTech Market RU for corporate training
+- **Mode:** Full Pipeline (/analyze)
+- **Session:** 3 of 5
+- **Previous Sessions:** 1 (Interview + Brief), 2 (Team Alpha: Research + Analysis + Strategy)
+
+### 2. Loading
+Read:
+- `domains/analytics/AGENTS.md`
+- `domains/analytics/.agents/skills/board/SKILL.md`
+- `domains/analytics/.agents/skills/handoff/SKILL.md`
+- `domains/analytics/.agents/skills/gates/SKILL.md`
+- `domains/analytics/.agents/skills/web-research/SKILL.md`
+- `domains/analytics/.agents/skills/competitive-analysis/SKILL.md`
+- `domains/analytics/.agents/skills/porters-five-forces/SKILL.md`
+
+Activate agents: Conductor (COND-03), Researcher (RES-02), Data Analyst (AN-02), Strategist (ST-02).
+Role: **Team Beta (Critics)** — task: critique and alternative analysis of Team Alpha's results.
+
+### 3. What is Completed
+**Session 1:** Conducted interview, formed research brief (12 questions), defined scope: corporate training, B2B SaaS, RU, 3-year horizon. Evaluation mode: scoring.
+**Session 2:** Team Alpha (Analysts) conducted research (28 sources), applied TAM/SAM/SOM + Competitive Analysis + PEST. Strategic recommendation: "AI-first corporate training platform, focus on 50-500 headcount companies."
+
+### 4. Data (full text)
+
+#### Research Brief (from INT-01)
+1. What is the size of EdTech Market RU (TAM/SAM/SOM) for corporate training?
+2. Who are the key competitors and what are their shares?
+3. What technological trends are changing the market?
+4. What regulatory factors affect EdTech?
+[... all 12 questions ...]
+
+#### Team Alpha Research Data (from RES-01)
+**TAM:** $4.2B (EdTech RU, 2026). Source: Smart Ranking, Jan 2026 ✅
+**SAM:** $1.8B (corporate training). Source: HSE, Mar 2026 ✅
+**SOM:** $180M (B2B SaaS platforms). Source: calculation ⚠️ (2 sources instead of 3)
+
+**Competitors:**
+| Competitor | Share | Threat Score | Key Strength |
+|-----------|------|-------------|---------------|
+| Skillbox (B2B) | 15% | 7.2 | Brand, content |
+| Netology | 12% | 6.8 | Corp. programs |
+| Yandex Praktikum | 18% | 7.6 | Yandex Ecosystem |
+[... full data ...]
+
+#### Team Alpha Analysis Results (from AN-01, ST-01)
+[FULL TEXT: TAM/SAM/SOM framework, Competitive Analysis, PEST, strategic recommendations]
+
+### 5. Current State
+
+#### Task Board
+📊 Progress: 40% (6/15) | Session: 3/5 | Blockers: 0
+
+| # | ID | Gate | Session | Status | Deliverable |
+|---|----|------|:------:|:------:|-------------|
+| 1 | COND-01 | Conductor | 1 | [✓] | Full Pipeline |
+| 2 | INT-01 | Interviewer | 1 | [✓] | Brief: 12 questions |
+| 3 | COND-02 | Conductor (s.2) | 2 | [✓] | Context loaded |
+| 4 | RES-01 | Researcher (Alpha) | 2 | [✓] | 28 sources |
+| 5 | AN-01 | Data Analyst (Alpha) | 2 | [✓] | 3 frameworks |
+| 6 | ST-01 | Strategist (Alpha) | 2 | [✓] | AI-first strategy |
+| 7 | COND-03 | Conductor (s.3) | 3 | [→] | — |
+| 8 | RES-02 | Researcher (Beta) | 3 | [ ] | — |
+| 9 | AN-02 | Data Analyst (Beta) | 3 | [ ] | — |
+| 10 | ST-02 | Strategist (Beta) | 3 | [ ] | — |
+[... remaining gates ...]
+
+#### Open Questions
+- ⚠️ SOM is calculated based on 2 sources (Alpha recommended verifying)
+
+### 6. Task for Session 3
+**Gates:** COND-03 → RES-02 → AN-02 → ST-02
+**What to do:** Team Beta (Critics) must:
+1. **RES-02:** Investigate the EdTech market from alternative perspectives. Check Alpha's data. Find counter-arguments. Verify SOM (find 3rd source).
+2. **AN-02:** Apply alternative frameworks (Porter's 5 Forces, Blue Ocean). Find weaknesses in Alpha's analysis.
+3. **ST-02:** Formulate an alternative strategy. Critique Alpha's strategy with reasoning.
+**Expected Result:** Alternative package (data + analysis + strategy) for handoff to Mediator in session 4.
+
+→ Start as Conductor (COND-03). Load the board, check Alpha's data, pass control to Researcher (RES-02).
+---
+```
+
+## Validation (Quality Gate)
+
+A prompt is considered correct if:
+
+- [ ] The prompt is self-sufficient — work can start without external context
+- [ ] Project name and mode are specified
+- [ ] Session number and session history are specified
+- [ ] Agents and skills to load are listed with file paths (from mapping)
+- [ ] All 🔴 Critical data is included as full text
+- [ ] Task board (`$board`) is included in its current state with metrics
+- [ ] The next gate and task are formulated specifically (gate, action, deliverable)
+- [ ] Open questions and blockers are noted
+- [ ] No forbidden elements (links without content, abbreviations, assumptions)
+- [ ] Validation test (7 points, step 4) — all "Yes"
+- [ ] In emergency mode — prompt is marked ⚠️, lost data is indicated
+
+## Handoff
+The output of `$session-prompt-generator` is an input for:
+- **Next session** — via user copying the prompt.
+- **`$handoff`** — the prompt is included in the `docs/analytics/session-N-handoff.md` file.
+- **`$board`** — the board state is saved in the prompt.
+
+Handoff format: prompt in markdown, ready to be pasted at the beginning of the new session.
+
+## Anti-patterns
+
+| Error | Why it's bad | How to do it right |
+|--------|-------------|---------------|
+| Prompt without research data | The next session will lose all collected facts | Include FULL text of data (🔴 Critical) |
+| "Continue work" | No context on what to continue | Specific task: gate, action, deliverable |
+| File links without content | Files might be inaccessible | Include content directly in the prompt |
+| Prompt without task board | Unclear where we are in the pipeline | Always include `$board` with metrics |
+| Abbreviated artifacts | Critical details are lost | Full text of every 🔴 Critical artifact |
+| Skipping open questions | Problems will resurface | Note all blockers and questions |
+| No loading instructions | The session will not load the needed agents | Mapping → specific file paths |
+| Prompt > 50% "fluff" | Dilution of critical data | Structure: blocks, tables, specifics |
+| Abnormal termination without prompt | Total loss of context | Emergency mode: minimal prompt + ⚠️ mark |
+
+## Output Template
+
+```markdown
+---
+## Prompt for Session [N+1] — [Project Name]
+
+### 1. Project
+- **Name:** [Name]
+- **Mode:** Full Pipeline (/analyze)
+- **Session:** [N+1] of 5
+- **Previous Sessions:** 1-[N] completed
+
+### 2. Loading
+Read:
+- `domains/analytics/AGENTS.md`
+- `domains/analytics/.agents/skills/board/SKILL.md`
+- `domains/analytics/.agents/skills/handoff/SKILL.md`
+- `domains/analytics/.agents/skills/gates/SKILL.md`
+- [Specific skills: file paths]
+
+Activate agents: [List of agents with Gate IDs]
+
+### 3. What is Completed
+**Session 1:** [Summary 2-3 sentences]
+**Session [N]:** [Summary 2-3 sentences]
+
+**Key decisions:**
+| # | Decision | Reasoning | Adopted by |
+|---|---------|-------------|------------|
+| 1 | [Decision] | [Why] | [User / Conductor] |
+
+### 4. Data (full text)
+
+#### Research Brief (from INT-01)
+[FULL TEXT research brief]
+
+#### Research Data (from RES-xx)
+[FULL TEXT data with sources and URLs]
+
+#### Analysis Results (from AN-xx)
+[FULL TEXT frameworks and findings]
+
+#### [Other artifacts as needed]
+[FULL TEXT]
+
+### 5. Current State
+
+#### Task Board
+[Full board from $board with progress metrics]
+
+#### Open Questions
+| # | Question / Blocker | Severity | For which gate |
+|---|-----------------|:--------:|:----------------:|
+| 1 | [Description] | 🔴/🟡/🟢 | [ID] |
+
+### 6. Task for Session [N+1]
+**Gates:** [ID list]
+**What to do:** [Specific description per gate]
+**Expected Result:** [Deliverables]
+**Constraints:** [If any]
+
+→ Start as Conductor (COND-xx). Load the board, check the data, pass control to [next agent].
+---
+```
+
+### Emergency Prompt (minimal version)
+
+```markdown
+---
+## ⚠️ EMERGENCY PROMPT — Session [N+1] — [Project Name]
+
+**Status:** Session [N] terminated abnormally. Possible data loss.
+
+### What is Saved
+[Everything collected: board, latest deliverable, questions]
+
+### What might be lost
+| # | Data | Recovery Method |
+|---|--------|-----------------------|
+| 1 | [Lost data] | [Repeated web-research / ask user / recover from handoff] |
+
+### Task
+Restore context from `docs/analytics/session-[N]-handoff.md`. Check data integrity. Continue from gate [ID].
+
+→ Start as Conductor (COND-xx). Priority: verify data integrity before continuing.
+---
+```