code-ai-installer 4.0.0 → 4.0.1-a

package/domains/analytics/locales/en/.agents/skills/gates/SKILL.md

@@ -1,366 +1,388 @@
----
-name: gates
-description: Quality control of analytics pipeline gates — passing criteria, checklists, blockers
----
-# Gates — Quality Control of Analytical Pipeline Gates
-
-## When to use
-- **Before every transition** between gates — to check the deliverable of the current gate.
-- **Upon return for revision** — for re-checking after fixes are made.
-- **At the end of a session** — to verify the completeness of materials prior to the inter-session `$handoff`.
-- **In case of readiness disputes** — as a formal arbitrator: what is done and what is not.
-
-> **Distinction from `$board`:** `$gates` checks **quality and completeness** of a deliverable. `$board` tracks **status and progress**. Gates answers "is it ready?", Board answers "where are we now?".
-
-## Input
-
-| Field | Required | Description |
-|-------|:--------:|-------------|
-| Gate Name | ✅ | Which gate is being checked (with ID: COND-01, INT-01, etc.) |
-| Pipeline Mode | ✅ | Full (`/analyze`) / Quick (`/quick-insight`) — determines the set of gates and criteria |
-| Gate Deliverables | ✅ | Artifacts that the agent transfers for review |
-| Iteration | ⬚ | If a repeat check — iteration number + gaps from the previous one |
-| Session Number | ⬚ | Current session (for Full Pipeline) |
-
-### Connection with other skills
-| Skill | Interaction |
-|-------|-------------|
-| `$board` | After gate check — update the gate status on the board (PASS → `[✓]`, FAIL → `[↩]`) |
-| `$handoff` | On PASS / CONDITIONAL PASS — form a handoff envelope with the gate check attached |
-| `$web-research` | Researcher gates (RES-xx): absence of web-research = automatic Blocker |
-| `$session-prompt-generator` | At the end of a session — the gate check goes into the data for the next session's prompt |
-| Framework skills | Data Analyst gates (AN-xx): check the correct completion of the Quality Gate for each framework |
-
-## Gates and Deliverables Registry
-
-### Full Pipeline (`/analyze`, 5 sessions)
-
-| # | Gate | ID | Agent | Session | Required deliverables | Skills |
-|---|------|----|-------|---------|-----------------------|--------|
-| 1 | **Conductor** | COND-01 | Conductor | 1 | Mode determined, task scope, board created | `$board` |
-| 2 | **Interviewer** | INT-01 | Interviewer | 1 | Context gathered, assessment mode selected (scoring/qualitative), research brief | `$handoff` |
-| 3 | **Researcher (Analysts)** | RES-01 | Researcher | 2 | Data collected, sources verified, minimum sources reached | `$web-research` |
-| 4 | **Data Analyst (Analysts)** | AN-01 | Data Analyst | 2 | Frameworks applied, methodology documented | Framework skills |
-| 5 | **Strategist (Analysts)** | ST-01 | Strategist | 2 | Recommendations actionable, linked to evidence | — |
-| 6 | **Researcher (Critics)** | RES-02 | Researcher | 3 | Alternative data, criticism of Analysts' data, sources verified | `$web-research` |
-| 7 | **Data Analyst (Critics)** | AN-02 | Data Analyst | 3 | Alternative analysis, counter-frameworks | Framework skills |
-| 8 | **Strategist (Critics)** | ST-02 | Strategist | 3 | Alternative strategy, criticism of Analysts' strategy | — |
-| 9 | **Mediator** | MED-01 | Mediator | 4 | Both teams evaluated, scoring/qualitative completed, final conclusion | `$gates` |
-| 10 | **Designer** | DS-01 | Designer | 5 | Report structure approved, visualization specifications | — |
-| 11 | **Layouter** | LY-01 | Layouter | 5 | HTML generated, formatting validated | — |
-| 12 | **Release Gate** | RG-01 | Conductor | 5 | All gates passed, PDF generated | `$board`, `$gates` |
-
-### Quick Pipeline (`/quick-insight`, 1 session)
-
-| # | Gate | ID | Agent | Required deliverables |
-|---|------|----|-------|-----------------------|
-| 1 | **Conductor** | COND-01 | Conductor | Mode determined, task scope, board created |
-| 2 | **Interviewer** | INT-01 | Interviewer | Context gathered, assessment mode selected |
-| 3 | **Researcher** | RES-01 | Researcher | Data collected, sources verified |
-| 4 | **Data Analyst** | AN-01 | Data Analyst | Frameworks applied |
-| 5 | **Strategist** | ST-01 | Strategist | Recommendations actionable |
-| 6 | **Mediator** | MED-01 | Mediator | Final conclusion (without competing teams) |
-| 7 | **Designer** | DS-01 | Designer | Report structure, visualizations |
-| 8 | **Layouter** | LY-01 | Layouter | HTML generated, PDF ready |
-
-## Readiness Criteria by Gates
-
-> Each criterion has a preset severity: **[B]** = Blocker, **[G]** = Gap, **[N]** = Note. During a check, severity can be upgraded (G→B), but not downgraded (B→G).
-
-### Conductor (COND-xx)
-- [ ] **[B]** Pipeline mode determined and justified (Full / Quick)
-- [ ] **[B]** Task scope formulated (what to analyze, why, limitations)
-- [ ] **[B]** Board created via `$board` with correct gate IDs
-- [ ] **[G]** First gate activated
-- [ ] **[G]** For continuing sessions: previous handoff loaded and verified
-
-### Interviewer (INT-xx)
-- [ ] **[B]** User context gathered (business question, industry, company size, goals)
-- [ ] **[B]** Assessment mode selected: scoring (quantitative) or qualitative
-- [ ] **[B]** Research brief formed with specific questions for research
-- [ ] **[G]** Key metrics / evaluation criteria defined
-- [ ] **[G]** Scope fixed — what is included and what is NOT included in the analysis
-
-### Researcher (RES-xx)
-- [ ] **[B]** Data collected for all questions in the research brief
-- [ ] **[B]** Web search performed via `$web-research` (mandatory; absence = automatic Blocker)
-- [ ] **[B]** Sources verified: every data point has a URL, date, credibility level
-- [ ] **[B]** No data without origin (any fact without a source = Blocker)
-- [ ] **[G]** Minimum sources reached (set in research brief, default ≥ 3 per thesis)
-- [ ] **[G]** Cross-check: 2+ sources for key data
-- [ ] **[G]** Conflicting data flagged and explained
-- [ ] **[G]** Data is up-to-date: market data — not older than 12 months, regulatory — current version
-- [ ] **[N]** Additional sources for depth (nice-to-have)
-
-**Dependency check:** Research brief from INT-xx must be explicitly cited. Each question of the brief = section in the deliverable.
-
-### Data Analyst (AN-xx)
-- [ ] **[B]** Frameworks chosen justifiably (at least 2 for Full Pipeline)
-- [ ] **[B]** Each framework applied correctly — all sections filled out, framework's Quality Gate passed
-- [ ] **[B]** Data from Researcher is linked to conclusions (not "in our opinion," but "according to data [source]")
-- [ ] **[G]** Methodology documented: why this framework, what data was used
-- [ ] **[G]** Patterns and insights explicitly highlighted
-- [ ] **[N]** Limitations of the analysis are recorded
-
-**Dependency check:** Deliverable from RES-xx must be explicitly used. Each key conclusion points to Researcher data, and not to own assumptions.
-
-### Strategist (ST-xx)
-- [ ] **[B]** Recommendations are actionable — specific actions, not generic advice
-- [ ] **[B]** Each recommendation is tied to evidence from Data Analyst / Researcher
-- [ ] **[G]** Recommendation prioritization (impact / effort or similar matrix)
-- [ ] **[G]** Risks are outlined for each recommendation
-- [ ] **[N]** Time horizons are specified (short-term / medium-term / long-term)
-
-**Dependency check:** Every recommendation traces back to an insight in AN-xx → data in RES-xx. "Recommendation without evidence" = Blocker.
-
-### Mediator (MED-xx)
-- [ ] **[B]** Results of both teams obtained (for Full Pipeline) or of the single team (Quick)
-- [ ] **[B]** Scoring or qualitative evaluation performed (according to the mode selected in INT-01)
-- [ ] **[B]** Final conclusion formulated with justification
-- [ ] **[G]** For scoring: criteria and weights defined, scores justified
-- [ ] **[G]** For qualitative: arguments from both sides contrasted, conflicts resolved
-- [ ] **[N]** Strengths of each team highlighted
-
-**Dependency check:** Deliverables ST-01 and ST-02 (Full) or ST-01 (Quick) are explicitly cited and contrasted.
-
-### Designer (DS-xx)
-- [ ] **[B]** Report structure approved (table of contents, sections, order)
-- [ ] **[B]** Visualization specifications defined (chart types, data, palette)
-- [ ] **[G]** Layout aligned with the content of the Mediator
-- [ ] **[N]** Branding / stylistics defined
-
-**Dependency check:** Report structure covers all sections from the MED-xx conclusion.
-
-### Layouter (LY-xx)
-- [ ] **[B]** HTML generated according to Designer's layout
-- [ ] **[B]** All visualizations are rendered
-- [ ] **[G]** Formatting validated (indents, fonts, tables)
-- [ ] **[G]** PDF generated and verified
-
-**Dependency check:** HTML matches DS-xx specification. All visualizations from the specification are present.
-
-### Release Gate (RG-xx)
-- [ ] **[B]** All previous gates passed (status `[✓]` on `$board`)
-- [ ] **[B]** PDF report generated and verified
-- [ ] **[G]** Data is up-to-date as of the report date
-- [ ] **[G]** User sign-off obtained
-
-## Criteria Severity
-
-| Severity | Marker | Definition | Impact on Decision |
-|----------|:------:|------------|--------------------|
-| Blocker | **[B]** | Mandatory deliverable is missing or critically incomplete | FAIL — return for revision |
-| Gap | **[G]** | Deliverable exists, but is incomplete or has remarks | CONDITIONAL PASS — can proceed, but gaps are logged |
-| Note | **[N]** | Minor remark, does not affect transition | PASS — remark is passed to the next gate as context |
-
-> Severity can be upgraded (G→B for critical issues), but **cannot be downgraded** (B→G). The preset severity is the minimum.
-
-## Decisions
-
-| Decision | Condition | Action |
-|----------|-----------|--------|
-| **PASS** | 0 Blockers + 0 Gaps | Request user Approved → `$handoff` → next gate |
-| **CONDITIONAL PASS** | 0 Blockers + 1-2 Gaps | Log gaps → Request user Approved with a note → `$handoff` with a note |
-| **FAIL** | ≥ 1 Blocker | Return to the agent with a specific list of gaps → Do not request Approved |
-
-> **Absolute Rule:** Never skip gates, even if the task seems simple. FAIL — revision — repeat check. Without exceptions.
-
-### Special Rules
-- Researcher gates (RES-xx): lack of `$web-research` — **automatic Blocker**.
-- Data without a source — **Blocker** (not Gap, not Note).
-- Data older than 12 months (market) — **Gap** with "requires update" note.
-
-## Escalation Protocol on Repeated FAILs
-
-| Iteration | Action |
-|:---------:|--------|
-| 1st FAIL | Standard return: list of gaps → agent revises → repeat check |
-| 2nd FAIL | Extended return: gaps + specific instructions "how to fix" + warning |
-| 3rd FAIL | **Escalation to User:** |
-
-Escalation Template (3+ FAILs):
-```
-### ⚠️ Escalation — Gate [ID] fails check (iteration [N])
-
-**Gate:** [ID] [Name]
-**FAIL Iterations:** [N]
-**Recurring gaps:** [list of gaps not fixed in N iterations]
-
-**Options:**
-1. Re-evaluate scope / lower deliverable requirements
-2. Return to the previous gate (problem is in the source data)
-3. Forced CONDITIONAL PASS (user accepts the risk of the gaps)
-
-→ Waiting for a decision from the user.
-```
-
-## Protocol
-
-### Step 0 — Loading Context
-1. Determine which gate is being checked (by ID from `$board`).
-2. Determine pipeline mode (Full / Quick) and session number.
-3. Load readiness criteria for this gate from the registry (with preset severity).
-4. If a repeat check — load gaps from the previous iteration.
-
-### Step 1 — Reviewing Deliverables
-1. For each readiness criterion — check: fulfilled / unfulfilled / N/A.
-2. For unfulfilled ones — apply the preset severity (can be upgraded, cannot be downgraded).
-3. For unfulfilled ones — describe the specific gap: what exactly is missing / what to fix.
-4. **Dependency check:** verify that the deliverable from the previous gate is explicitly used (not merely existing, but cited/utilized in the current deliverable).
-
-### Step 2 — Decision Making
-1. Calculate severity.
-2. Apply decision rules (PASS / CONDITIONAL PASS / FAIL).
-3. Verify iteration number — if 3+ FAILs, activate escalation protocol.
-4. Add entry to check log.
-
-### Step 3 — Action by Result
-- **PASS:** show to the user → request Approved → `$handoff` → update `$board`.
-- **CONDITIONAL PASS:** show gaps → request Approved acknowledging gaps → `$handoff` with note → update `$board`.
-- **FAIL:** show gaps → return to the agent → update `$board` (status `[↩] Return`). If 3+ FAILs — escalate.
-
-## Example — Gate Check RES-01 (Researcher Analysts), iteration 1 → CONDITIONAL PASS
-
-```
-### Gate Check — RES-01 Researcher (Analysts)
-
-**Mode:** Full Pipeline
-**Session:** 2 of 5
-**Iteration:** 1
-**Previous gaps:** Initial check
-
----
-
-#### Readiness Criteria
-
-| # | Criterion | Sev | Status | Gap / Comment |
-|---|-----------|:---:|:------:|---------------|
-| 1 | Data collected for all questions in the research brief | [B] | ✅ | 7 out of 7 questions covered |
-| 2 | Web search performed via `$web-research` | [B] | ✅ | 12 queries, 28 sources |
-| 3 | Sources verified (URL, date, credibility) | [B] | ✅ | All 28 have URL and date |
-| 4 | No data without origin | [B] | ✅ | 0 facts without source |
-| 5 | Minimum sources (≥ 3 per thesis) | [G] | ⚠️ | EdTech market TAM thesis — 2 sources instead of 3 |
-| 6 | Cross-check (2+ for key data) | [G] | ✅ | OK |
-| 7 | Conflicting data flagged | [G] | ✅ | 2 conflicts flagged and explained |
-| 8 | Data is up-to-date (≤ 12 months) | [G] | ⚠️ | 1 source (IDC report) — 14 mo, flagged ⚠️ |
-| 9 | Additional sources (nice-to-have) | [N] | ℹ️ | Recommended to add Rosstat data |
-
-**Dependency check:** Research brief INT-01 cited in the heading of each section. All 7 questions = 7 sections. ✅
-
----
-
-#### Summary
-
-| Severity | Quantity | Items |
-|----------|:--------:|-------|
-| Blocker | 0 | — |
-| Gap | 2 | #5 (TAM: 2 sources), #8 (IDC: 14 mo) |
-| Note | 1 | #9 (Rosstat nice-to-have) |
-
-### Decision: CONDITIONAL PASS
-
-**Justification:** 0 Blockers, 2 Gaps — acceptable to proceed. Gaps logged for AN-01.
-
----
-
-**Deliverables to transfer:**
-- Research Report (7 sections, 28 sources, 2 conflicts explained)
-
-**Gaps for AN-01:**
-- EdTech market TAM supported by 2 sources (instead of 3) — note as ⚠️ Estimated
-- IDC report 14 months old — flagged, recommended to verify with fresh data
-
-→ Ready to handoff to **Data Analyst (Analysts)**. Waiting for **"Approved"**.
-```
-
-## Validation (Quality Gate)
-
-The gate check is considered complete if:
-
-- [ ] All readiness criteria have been checked (fulfilled / unfulfilled / N/A — no blanks)
-- [ ] Each criterion is assigned severity from the registry (B/G/N), upgraded if necessary
-- [ ] For every unfulfilled item specified: severity, gap description, what to fix
-- [ ] Dependency check passed: the deliverable of the previous gate is explicitly used in the current one
-- [ ] Decision conforms to the rules (not subjective)
-- [ ] On 3+ FAILs — escalation protocol activated
-- [ ] Check log updated
-- [ ] On PASS/CONDITIONAL PASS — user Approved obtained
-- [ ] `$board` updated
-
-## Handoff
-The result of `$gates` is input data for: `$handoff` (attaching the gate check to the envelope).
-
-Transfer format: filled gate check template + decision + gaps (if CONDITIONAL PASS) + dependency check result.
-
-## Anti-patterns
-
-| Mistake | Why it's bad | How to do it right |
-|---------|--------------|--------------------|
-| Skipping a gate "because it's simple" | Errors go unnoticed | Never skip. Without exceptions |
-| PASS without checking all criteria | Unnoticed gaps | Every criterion = fulfilled / unfulfilled / N/A |
-| "Overall ok" instead of specifics | The agent doesn't know what to fix | Specific gap: "source X does not contain a publication date" |
-| FAIL without describing gaps | The agent returns and doesn't know what to do | Every FAIL = list of specific fixes |
-| Approved without user sign-off | Pipeline protocol violation | Always an explicit Approved from the user |
-| Researcher data without web search | Unverified data in the report | `$web-research` — mandatory, absence is a Blocker |
-| Equal importance of all criteria | Minor issue blocks, critical issue passes | Preset severity: [B] / [G] / [N] per criterion |
-| Repeat check without comparing gaps | A gap might not be fixed | Load gaps from previous iteration, check each one |
-| Downgrading severity (B→G) | Critical deliverable passes without fixes | Severity can be upgraded, but not downgraded |
-| Endless FAIL loop | Pipeline hangs, user is unaware | 3+ FAILs = escalation to the user with options to resolve |
-| Formal check "it exists" | Deliverable is there, but not used by the next gate | Dependency check: previous deliverable is cited and put to use |
-
-## Output Template
-
-```
-### Gate Check — [ID] [Gate Name]
-
-**Mode:** [Full / Quick]
-**Session:** [N out of M]
-**Iteration:** [1 / 2 / N]
-**Previous gaps:** [list or "Initial check"]
-
----
-
-#### Readiness Criteria
-
-| # | Criterion | Sev | Status | Gap / Comment |
-|---|-----------|:---:|:------:|---------------|
-| 1 | [Criterion from registry] | [B] | ✅/⚠️/❌ | [If unfulfilled: what is missing] |
-
-**Dependency check:** [Deliverable [previous ID] cited: ✅/❌. Details.]
-
----
-
-#### Summary
-
-| Severity | Quantity | Items |
-|----------|:--------:|-------|
-| Blocker | [N] | #[list] |
-| Gap | [N] | #[list] |
-| Note | [N] | #[list] |
-
-### Decision: [ PASS / CONDITIONAL PASS / FAIL ]
-
-**Justification:** [per rules: 0 blockers + 0 gaps = PASS, etc.]
-
----
-
-#### If PASS / CONDITIONAL PASS:
-**Deliverables to transfer:**
-- [Artifact 1]
-- [Artifact 2]
-
-**Gaps for next gate (if CONDITIONAL):**
-- [Gap 1 — what to consider]
-
-→ Ready to handoff to **[Next Agent]**. Waiting for **"Approved"**.
-
-#### If FAIL:
-**Gaps for revision:**
-
-| # | Gap | Sev | What to fix | Responsible |
-|---|-----|:---:|-------------|-------------|
-| [N] | [description] | [B] | [specific action] | [agent] |
-
-→ Return to **[Current Agent]**. After fixing — repeat check (iteration [N+1]).
-[If iteration ≥ 3: escalation protocol activated]
-```
+---
+name: gates
+description: Quality control of analytics pipeline gates — passing criteria, checklists, blockers
+type: mandatory
+domain: analytics
+owners:
+  - conductor
+gates:
+  - INTERVIEWER
+  - RESEARCHER
+  - DATA_ANALYST
+  - STRATEGIST
+  - MEDIATOR
+  - DESIGNER
+  - LAYOUTER
+  - RELEASE_GATE
+tech: []
+topic: []
+triggers:
+  - "gates"
+  - "Контроль гейтов аналитического пайплайна"
+  - "Контроль качества гейтов аналитического пайплайна — критерии прохождения, чек-листы, блокеры"
+related: []
+budget_lines: 388
+schema_version: 1
+---
+# Gates — Quality Control of Analytical Pipeline Gates
+
+## When to use
+- **Before every transition** between gates — to check the deliverable of the current gate.
+- **Upon return for revision** — for re-checking after fixes are made.
+- **At the end of a session** — to verify the completeness of materials prior to the inter-session `$handoff`.
+- **In case of readiness disputes** — as a formal arbitrator: what is done and what is not.
+
+> **Distinction from `$board`:** `$gates` checks **quality and completeness** of a deliverable. `$board` tracks **status and progress**. Gates answers "is it ready?", Board answers "where are we now?".
+
+## Input
+
+| Field | Required | Description |
+|-------|:--------:|-------------|
+| Gate Name | ✅ | Which gate is being checked (with ID: COND-01, INT-01, etc.) |
+| Pipeline Mode | ✅ | Full (`/analyze`) / Quick (`/quick-insight`) — determines the set of gates and criteria |
+| Gate Deliverables | ✅ | Artifacts that the agent transfers for review |
+| Iteration | ⬚ | If a repeat check — iteration number + gaps from the previous one |
+| Session Number | ⬚ | Current session (for Full Pipeline) |
+
+### Connection with other skills
+| Skill | Interaction |
+|-------|-------------|
+| `$board` | After gate check — update the gate status on the board (PASS → `[✓]`, FAIL → `[↩]`) |
+| `$handoff` | On PASS / CONDITIONAL PASS — form a handoff envelope with the gate check attached |
+| `$web-research` | Researcher gates (RES-xx): absence of web-research = automatic Blocker |
+| `$session-prompt-generator` | At the end of a session — the gate check goes into the data for the next session's prompt |
+| Framework skills | Data Analyst gates (AN-xx): check the correct completion of the Quality Gate for each framework |
+
+## Gates and Deliverables Registry
+
+### Full Pipeline (`/analyze`, 5 sessions)
+
+| # | Gate | ID | Agent | Session | Required deliverables | Skills |
+|---|------|----|-------|---------|-----------------------|--------|
+| 1 | **Conductor** | COND-01 | Conductor | 1 | Mode determined, task scope, board created | `$board` |
+| 2 | **Interviewer** | INT-01 | Interviewer | 1 | Context gathered, assessment mode selected (scoring/qualitative), research brief | `$handoff` |
+| 3 | **Researcher (Analysts)** | RES-01 | Researcher | 2 | Data collected, sources verified, minimum sources reached | `$web-research` |
+| 4 | **Data Analyst (Analysts)** | AN-01 | Data Analyst | 2 | Frameworks applied, methodology documented | Framework skills |
+| 5 | **Strategist (Analysts)** | ST-01 | Strategist | 2 | Recommendations actionable, linked to evidence | — |
+| 6 | **Researcher (Critics)** | RES-02 | Researcher | 3 | Alternative data, criticism of Analysts' data, sources verified | `$web-research` |
+| 7 | **Data Analyst (Critics)** | AN-02 | Data Analyst | 3 | Alternative analysis, counter-frameworks | Framework skills |
+| 8 | **Strategist (Critics)** | ST-02 | Strategist | 3 | Alternative strategy, criticism of Analysts' strategy | — |
+| 9 | **Mediator** | MED-01 | Mediator | 4 | Both teams evaluated, scoring/qualitative completed, final conclusion | `$gates` |
+| 10 | **Designer** | DS-01 | Designer | 5 | Report structure approved, visualization specifications | — |
+| 11 | **Layouter** | LY-01 | Layouter | 5 | HTML generated, formatting validated | — |
+| 12 | **Release Gate** | RG-01 | Conductor | 5 | All gates passed, PDF generated | `$board`, `$gates` |
+
+### Quick Pipeline (`/quick-insight`, 1 session)
+
+| # | Gate | ID | Agent | Required deliverables |
+|---|------|----|-------|-----------------------|
+| 1 | **Conductor** | COND-01 | Conductor | Mode determined, task scope, board created |
+| 2 | **Interviewer** | INT-01 | Interviewer | Context gathered, assessment mode selected |
+| 3 | **Researcher** | RES-01 | Researcher | Data collected, sources verified |
+| 4 | **Data Analyst** | AN-01 | Data Analyst | Frameworks applied |
+| 5 | **Strategist** | ST-01 | Strategist | Recommendations actionable |
+| 6 | **Mediator** | MED-01 | Mediator | Final conclusion (without competing teams) |
+| 7 | **Designer** | DS-01 | Designer | Report structure, visualizations |
+| 8 | **Layouter** | LY-01 | Layouter | HTML generated, PDF ready |
+
+## Readiness Criteria by Gates
+
+> Each criterion has a preset severity: **[B]** = Blocker, **[G]** = Gap, **[N]** = Note. During a check, severity can be upgraded (G→B), but not downgraded (B→G).
+
+### Conductor (COND-xx)
+- [ ] **[B]** Pipeline mode determined and justified (Full / Quick)
+- [ ] **[B]** Task scope formulated (what to analyze, why, limitations)
+- [ ] **[B]** Board created via `$board` with correct gate IDs
+- [ ] **[G]** First gate activated
+- [ ] **[G]** For continuing sessions: previous handoff loaded and verified
+
+### Interviewer (INT-xx)
+- [ ] **[B]** User context gathered (business question, industry, company size, goals)
+- [ ] **[B]** Assessment mode selected: scoring (quantitative) or qualitative
+- [ ] **[B]** Research brief formed with specific questions for research
+- [ ] **[G]** Key metrics / evaluation criteria defined
+- [ ] **[G]** Scope fixed — what is included and what is NOT included in the analysis
+
+### Researcher (RES-xx)
+- [ ] **[B]** Data collected for all questions in the research brief
+- [ ] **[B]** Web search performed via `$web-research` (mandatory; absence = automatic Blocker)
+- [ ] **[B]** Sources verified: every data point has a URL, date, credibility level
+- [ ] **[B]** No data without origin (any fact without a source = Blocker)
+- [ ] **[G]** Minimum sources reached (set in research brief, default ≥ 3 per thesis)
+- [ ] **[G]** Cross-check: 2+ sources for key data
+- [ ] **[G]** Conflicting data flagged and explained
+- [ ] **[G]** Data is up-to-date: market data — not older than 12 months, regulatory — current version
+- [ ] **[N]** Additional sources for depth (nice-to-have)
+
+**Dependency check:** Research brief from INT-xx must be explicitly cited. Each question of the brief = section in the deliverable.
+
+### Data Analyst (AN-xx)
+- [ ] **[B]** Frameworks chosen justifiably (at least 2 for Full Pipeline)
+- [ ] **[B]** Each framework applied correctly — all sections filled out, framework's Quality Gate passed
+- [ ] **[B]** Data from Researcher is linked to conclusions (not "in our opinion," but "according to data [source]")
+- [ ] **[G]** Methodology documented: why this framework, what data was used
+- [ ] **[G]** Patterns and insights explicitly highlighted
+- [ ] **[N]** Limitations of the analysis are recorded
+
+**Dependency check:** Deliverable from RES-xx must be explicitly used. Each key conclusion points to Researcher data, and not to own assumptions.
+
+### Strategist (ST-xx)
+- [ ] **[B]** Recommendations are actionable — specific actions, not generic advice
+- [ ] **[B]** Each recommendation is tied to evidence from Data Analyst / Researcher
+- [ ] **[G]** Recommendation prioritization (impact / effort or similar matrix)
+- [ ] **[G]** Risks are outlined for each recommendation
+- [ ] **[N]** Time horizons are specified (short-term / medium-term / long-term)
+
+**Dependency check:** Every recommendation traces back to an insight in AN-xx → data in RES-xx. "Recommendation without evidence" = Blocker.
+
+### Mediator (MED-xx)
+- [ ] **[B]** Results of both teams obtained (for Full Pipeline) or of the single team (Quick)
+- [ ] **[B]** Scoring or qualitative evaluation performed (according to the mode selected in INT-01)
+- [ ] **[B]** Final conclusion formulated with justification
+- [ ] **[G]** For scoring: criteria and weights defined, scores justified
+- [ ] **[G]** For qualitative: arguments from both sides contrasted, conflicts resolved
+- [ ] **[N]** Strengths of each team highlighted
+
+**Dependency check:** Deliverables ST-01 and ST-02 (Full) or ST-01 (Quick) are explicitly cited and contrasted.
+
+### Designer (DS-xx)
+- [ ] **[B]** Report structure approved (table of contents, sections, order)
+- [ ] **[B]** Visualization specifications defined (chart types, data, palette)
+- [ ] **[G]** Layout aligned with the content of the Mediator
+- [ ] **[N]** Branding / stylistics defined
+
+**Dependency check:** Report structure covers all sections from the MED-xx conclusion.
+
+### Layouter (LY-xx)
+- [ ] **[B]** HTML generated according to Designer's layout
+- [ ] **[B]** All visualizations are rendered
+- [ ] **[G]** Formatting validated (indents, fonts, tables)
+- [ ] **[G]** PDF generated and verified
+
+**Dependency check:** HTML matches DS-xx specification. All visualizations from the specification are present.
+
+### Release Gate (RG-xx)
+- [ ] **[B]** All previous gates passed (status `[✓]` on `$board`)
+- [ ] **[B]** PDF report generated and verified
+- [ ] **[G]** Data is up-to-date as of the report date
+- [ ] **[G]** User sign-off obtained
+
+## Criteria Severity
+
+| Severity | Marker | Definition | Impact on Decision |
+|----------|:------:|------------|--------------------|
+| Blocker | **[B]** | Mandatory deliverable is missing or critically incomplete | FAIL — return for revision |
+| Gap | **[G]** | Deliverable exists, but is incomplete or has remarks | CONDITIONAL PASS — can proceed, but gaps are logged |
+| Note | **[N]** | Minor remark, does not affect transition | PASS — remark is passed to the next gate as context |
+
+> Severity can be upgraded (G→B for critical issues), but **cannot be downgraded** (B→G). The preset severity is the minimum.
+
+## Decisions
+
+| Decision | Condition | Action |
+|----------|-----------|--------|
+| **PASS** | 0 Blockers + 0 Gaps | Request user Approved → `$handoff` → next gate |
+| **CONDITIONAL PASS** | 0 Blockers + 1-2 Gaps | Log gaps → Request user Approved with a note → `$handoff` with a note |
+| **FAIL** | ≥ 1 Blocker | Return to the agent with a specific list of gaps → Do not request Approved |
+
+> **Absolute Rule:** Never skip gates, even if the task seems simple. FAIL — revision — repeat check. Without exceptions.
+
+### Special Rules
+- Researcher gates (RES-xx): lack of `$web-research` — **automatic Blocker**.
+- Data without a source — **Blocker** (not Gap, not Note).
+- Data older than 12 months (market) — **Gap** with "requires update" note.
+
+## Escalation Protocol on Repeated FAILs
+
+| Iteration | Action |
+|:---------:|--------|
+| 1st FAIL | Standard return: list of gaps → agent revises → repeat check |
+| 2nd FAIL | Extended return: gaps + specific instructions "how to fix" + warning |
+| 3rd FAIL | **Escalation to User:** |
+
+Escalation Template (3+ FAILs):
+```
+### ⚠️ Escalation — Gate [ID] fails check (iteration [N])
+
+**Gate:** [ID] [Name]
+**FAIL Iterations:** [N]
+**Recurring gaps:** [list of gaps not fixed in N iterations]
+
+**Options:**
+1. Re-evaluate scope / lower deliverable requirements
+2. Return to the previous gate (problem is in the source data)
+3. Forced CONDITIONAL PASS (user accepts the risk of the gaps)
+
+→ Waiting for a decision from the user.
+```
+
+## Protocol
+
+### Step 0 — Loading Context
+1. Determine which gate is being checked (by ID from `$board`).
+2. Determine pipeline mode (Full / Quick) and session number.
+3. Load readiness criteria for this gate from the registry (with preset severity).
+4. If a repeat check — load gaps from the previous iteration.
+
+### Step 1 — Reviewing Deliverables
+1. For each readiness criterion — check: fulfilled / unfulfilled / N/A.
+2. For unfulfilled ones — apply the preset severity (can be upgraded, cannot be downgraded).
+3. For unfulfilled ones — describe the specific gap: what exactly is missing / what to fix.
+4. **Dependency check:** verify that the deliverable from the previous gate is explicitly used (not merely existing, but cited/utilized in the current deliverable).
+
+### Step 2 — Decision Making
+1. Calculate severity.
+2. Apply decision rules (PASS / CONDITIONAL PASS / FAIL).
+3. Verify iteration number — if 3+ FAILs, activate escalation protocol.
+4. Add entry to check log.
+
+### Step 3 — Action by Result
+- **PASS:** show to the user → request Approved → `$handoff` → update `$board`.
+- **CONDITIONAL PASS:** show gaps → request Approved acknowledging gaps → `$handoff` with note → update `$board`.
+- **FAIL:** show gaps → return to the agent → update `$board` (status `[↩] Return`). If 3+ FAILs — escalate.
+
+## Example — Gate Check RES-01 (Researcher Analysts), iteration 1 → CONDITIONAL PASS
+
+```
+### Gate Check — RES-01 Researcher (Analysts)
+
+**Mode:** Full Pipeline
+**Session:** 2 of 5
+**Iteration:** 1
+**Previous gaps:** Initial check
+
+---
+
+#### Readiness Criteria
+
+| # | Criterion | Sev | Status | Gap / Comment |
+|---|-----------|:---:|:------:|---------------|
+| 1 | Data collected for all questions in the research brief | [B] | ✅ | 7 out of 7 questions covered |
+| 2 | Web search performed via `$web-research` | [B] | ✅ | 12 queries, 28 sources |
+| 3 | Sources verified (URL, date, credibility) | [B] | ✅ | All 28 have URL and date |
+| 4 | No data without origin | [B] | ✅ | 0 facts without source |
+| 5 | Minimum sources (≥ 3 per thesis) | [G] | ⚠️ | EdTech market TAM thesis — 2 sources instead of 3 |
+| 6 | Cross-check (2+ for key data) | [G] | ✅ | OK |
+| 7 | Conflicting data flagged | [G] | ✅ | 2 conflicts flagged and explained |
+| 8 | Data is up-to-date (≤ 12 months) | [G] | ⚠️ | 1 source (IDC report) — 14 mo, flagged ⚠️ |
+| 9 | Additional sources (nice-to-have) | [N] | ℹ️ | Recommended to add Rosstat data |
+
+**Dependency check:** Research brief INT-01 cited in the heading of each section. All 7 questions = 7 sections. ✅
+
+---
+
+#### Summary
+
+| Severity | Quantity | Items |
+|----------|:--------:|-------|
+| Blocker | 0 | — |
+| Gap | 2 | #5 (TAM: 2 sources), #8 (IDC: 14 mo) |
+| Note | 1 | #9 (Rosstat nice-to-have) |
+
+### Decision: CONDITIONAL PASS
+
+**Justification:** 0 Blockers, 2 Gaps — acceptable to proceed. Gaps logged for AN-01.
+
+---
+
+**Deliverables to transfer:**
+- Research Report (7 sections, 28 sources, 2 conflicts explained)
+
+**Gaps for AN-01:**
+- EdTech market TAM supported by 2 sources (instead of 3) — note as ⚠️ Estimated
+- IDC report 14 months old — flagged, recommended to verify with fresh data
+
+→ Ready to handoff to **Data Analyst (Analysts)**. Waiting for **"Approved"**.
+```
+
+## Validation (Quality Gate)
+
+The gate check is considered complete if:
+
+- [ ] All readiness criteria have been checked (fulfilled / unfulfilled / N/A — no blanks)
+- [ ] Each criterion is assigned severity from the registry (B/G/N), upgraded if necessary
+- [ ] For every unfulfilled item specified: severity, gap description, what to fix
+- [ ] Dependency check passed: the deliverable of the previous gate is explicitly used in the current one
+- [ ] Decision conforms to the rules (not subjective)
+- [ ] On 3+ FAILs — escalation protocol activated
+- [ ] Check log updated
+- [ ] On PASS/CONDITIONAL PASS — user Approved obtained
+- [ ] `$board` updated
+
+## Handoff
+The result of `$gates` is input data for: `$handoff` (attaching the gate check to the envelope).
+
+Transfer format: filled gate check template + decision + gaps (if CONDITIONAL PASS) + dependency check result.
+
+## Anti-patterns
+
+| Mistake | Why it's bad | How to do it right |
+|---------|--------------|--------------------|
+| Skipping a gate "because it's simple" | Errors go unnoticed | Never skip. Without exceptions |
+| PASS without checking all criteria | Unnoticed gaps | Every criterion = fulfilled / unfulfilled / N/A |
+| "Overall ok" instead of specifics | The agent doesn't know what to fix | Specific gap: "source X does not contain a publication date" |
+| FAIL without describing gaps | The agent returns and doesn't know what to do | Every FAIL = list of specific fixes |
+| Approved without user sign-off | Pipeline protocol violation | Always an explicit Approved from the user |
+| Researcher data without web search | Unverified data in the report | `$web-research` — mandatory, absence is a Blocker |
+| Equal importance of all criteria | Minor issue blocks, critical issue passes | Preset severity: [B] / [G] / [N] per criterion |
+| Repeat check without comparing gaps | A gap might not be fixed | Load gaps from previous iteration, check each one |
+| Downgrading severity (B→G) | Critical deliverable passes without fixes | Severity can be upgraded, but not downgraded |
+| Endless FAIL loop | Pipeline hangs, user is unaware | 3+ FAILs = escalation to the user with options to resolve |
+| Formal check "it exists" | Deliverable is there, but not used by the next gate | Dependency check: previous deliverable is cited and put to use |
+
+## Output Template
+
+```
+### Gate Check — [ID] [Gate Name]
+
+**Mode:** [Full / Quick]
+**Session:** [N out of M]
+**Iteration:** [1 / 2 / N]
+**Previous gaps:** [list or "Initial check"]
+
+---
+
+#### Readiness Criteria
+
+| # | Criterion | Sev | Status | Gap / Comment |
+|---|-----------|:---:|:------:|---------------|
+| 1 | [Criterion from registry] | [B] | ✅/⚠️/❌ | [If unfulfilled: what is missing] |
+
+**Dependency check:** [Deliverable [previous ID] cited: ✅/❌. Details.]
+
+---
+
+#### Summary
+
+| Severity | Quantity | Items |
+|----------|:--------:|-------|
+| Blocker | [N] | #[list] |
+| Gap | [N] | #[list] |
+| Note | [N] | #[list] |
+
+### Decision: [ PASS / CONDITIONAL PASS / FAIL ]
+
+**Justification:** [per rules: 0 blockers + 0 gaps = PASS, etc.]
+
+---
+
+#### If PASS / CONDITIONAL PASS:
+**Deliverables to transfer:**
+- [Artifact 1]
+- [Artifact 2]
+
+**Gaps for next gate (if CONDITIONAL):**
+- [Gap 1 — what to consider]
+
+→ Ready to handoff to **[Next Agent]**. Waiting for **"Approved"**.
+
+#### If FAIL:
+**Gaps for revision:**
+
+| # | Gap | Sev | What to fix | Responsible |
+|---|-----|:---:|-------------|-------------|
+| [N] | [description] | [B] | [specific action] | [agent] |
+
+→ Return to **[Current Agent]**. After fixing — repeat check (iteration [N+1]).
+[If iteration ≥ 3: escalation protocol activated]
+```