tools-only WORKFLOW.md - Orchestration Contracts
Version: 3.0.0 (20260201a) Purpose: Phase orchestration, structured data contracts, validation gates, gating protocol
⚠️ CRITICAL: Data Flow Architecture
PRINCIPLE: Markdown 是报告，YAML/JSON 是数据。两者必须分离！
┌─────────────────────────────────────────────────────────────────────────┐
│ STRICT DATA FLOW ARCHITECTURE │
├─────────────────────────────────────────────────────────────────────────┤
│ │
│ .phase_working/ │
│ ├── {SESSION_ID}/ ← Session isolation directory │
│ │ ├── _session_meta.yaml ← Session state (machine-readable) │
│ │ ├── data/ ← STRUCTURED DATA (for inter-phase) │
│ │ │ ├── P1_project_context.yaml │
│ │ │ ├── P2_dfd_elements.yaml │
│ │ │ ├── P3_boundary_context.yaml │
│ │ │ ├── P4_security_gaps.yaml │
│ │ │ ├── P5_threat_inventory.yaml │
│ │ │ ├── P6_validated_risks.yaml │
│ │ │ ├── P7_mitigation_plan.yaml │
│ │ │ └── P8_report_manifest.yaml │
│ │ └── reports/ ← HUMAN-READABLE REPORTS │
│ │ ├── P1-PROJECT-UNDERSTANDING.md │
│ │ ├── P2-DFD-ANALYSIS.md │
│ │ ├── P3-TRUST-BOUNDARY.md │
│ │ ├── P4-SECURITY-DESIGN-REVIEW.md │
│ │ ├── P5-STRIDE-THREATS.md │
│ │ ├── P6-RISK-VALIDATION.md │
│ │ └── P7-MITIGATION-PLAN.md │
│ └── _sessions_index.yaml ← All sessions index (optional) │
│ │
│ SESSION_ID FORMAT: {PROJECT_NAME}_{YYYYMMDD_HHMMSS} │
│ EXAMPLE: OPEN-WEBUI_20260130_143022 │
│ │
│ DATA FLOW: │
│ P1 → writes P1_project_context.yaml → P2 reads it │
│ P2 → writes P2_dfd_elements.yaml → P3 reads it │
│ ... (strict chain, no gaps) │
│ │
│ ❌ FORBIDDEN: P2 reading P1's .md report for data extraction │
│ ✅ REQUIRED: P2 reading P1's .yaml data file directly │
│ │
└─────────────────────────────────────────────────────────────────────────┘
§1 Session Initialization
Session ID Generation
SESSION_ID = {PROJECT_NAME}_{YYYYMMDD_HHMMSS}
Examples:
OPEN-WEBUI_20260130_143022
N8N_20260130_150000
DIFY_20260131_091530
Directory Creation
At session start, create:
SESSION_ID="${PROJECT_NAME}_$(date +%Y%m%d_%H%M%S)"
mkdir -p ".phase_working/${SESSION_ID}/data"
mkdir -p ".phase_working/${SESSION_ID}/reports"
Session Metadata
# .phase_working/{SESSION_ID}/_session_meta.yaml
schema_version: "3.0.0 (20260201a)"
session_id: "OPEN-WEBUI_20260130_143022"
project_name: "OPEN-WEBUI"
project_path: "/absolute/path/to/project"
started_at: "2026-01-30T14:30:22Z"
language: "en" # en|zh|ja|ko - Controls report output language
skill_version: "3.0.0 (20260201a)"
# Language Field Behavior:
# - "en": English reports, terminology, and recommendations
# - "zh": Chinese (Simplified) reports
# - "ja": Japanese reports
# - "ko": Korean reports
# Note: YAML data files always use English keys; language only affects
# human-readable reports (.md files) and recommendation text.
phases:
P1:
status: "pending" # pending|in_progress|completed|failed
started_at: null
completed_at: null
data_file: null # Path to P1_project_context.yaml when done
report_file: null # Path to P1-*.md when done
validation:
exit_code: null
errors: []
P2:
status: "pending"
# ... same structure
# P3-P8: same structure
Sessions Index (Optional)
# .phase_working/_sessions_index.yaml
sessions:
- session_id: "OPEN-WEBUI_20260130_143022"
project_name: "OPEN-WEBUI"
started_at: "2026-01-30T14:30:22Z"
status: "completed"
current_phase: 8
- session_id: "OPEN-WEBUI_20260129_100000"
project_name: "OPEN-WEBUI"
started_at: "2026-01-29T10:00:00Z"
status: "completed"
current_phase: 8
Todo Creation
Create 8 items at session start:
[
{"content": "Phase 1: Project Understanding", "status": "pending"},
{"content": "Phase 2: Call Flow & DFD Analysis", "status": "pending"},
{"content": "Phase 3: Trust Boundary Evaluation", "status": "pending"},
{"content": "Phase 4: Security Design Review", "status": "pending"},
{"content": "Phase 5: STRIDE Threat Analysis", "status": "pending"},
{"content": "Phase 6: Risk Validation", "status": "pending"},
{"content": "Phase 7: Mitigation Planning", "status": "pending"},
{"content": "Phase 8: Report Generation", "status": "pending"}
]
§2 Phase Execution Protocol (with Gating)
REFERENCE: 核心执行规则见 SKILL.md §10 Core Execution Rules
Per-Phase Workflow (Enhanced with Gating Protocol)
FOR each phase N in [1..8]:
┌─ ① PLANNING ───────────────────────────────────────────────────────────┐
│ │
│ 1. STATE PHASE OBJECTIVE │
│ - Explicitly declare: "Phase {N} objective is..." │
│ - Verify alignment with project goal │
│ │
│ 2. LOAD & VERIFY INPUT DATA (except P1) │
│ - Read .phase_working/data/P{N-1}_*.yaml │
│ - Parse as structured data │
│ - ❌ DO NOT read previous .md reports for data │
│ - ⚠️ STOP if input incomplete or malformed │
│ │
│ 3. DECOMPOSE INTO SUB-TASKS │
│ - Break phase work into 3-7 sub-tasks │
│ - Define input/output for each sub-task │
│ - Create todos for tracking │
│ │
└─────────────────────────────────────────────────────────────────────────┘
↓
┌─ ② EXECUTION LOOP ─────────────────────────────────────────────────────┐
│ │
│ 4. UPDATE SESSION META │
│ - Set phases.P{N}.status = "in_progress" │
│ - Set phases.P{N}.started_at = now() │
│ │
│ 5. EXECUTE ANALYSIS (per sub-task) │
│ - Follow @phases/P{N}-*.md instructions │
│ - Query knowledge base as needed (log queries!) │
│ - After each sub-task: verify alignment │
│ - ⚠️ If issue found: STOP → fix → verify → continue │
│ │
│ 6. WRITE OUTPUT DATA (YAML first!) │
│ - Write .phase_working/data/P{N}_*.yaml │
│ - Validate schema compliance │
│ - ✅ This is the PRIMARY output │
│ │
│ 7. WRITE REPORT (MD second) │
│ - Write .phase_working/reports/P{N}-*.md │
│ - Reference data from YAML, format for humans │
│ - ❌ This is SECONDARY, for human reading only │
│ │
└─────────────────────────────────────────────────────────────────────────┘
↓
┌─ ③ REFLECTION ─────────────────────────────────────────────────────────┐
│ │
│ 8. VALIDATE & UPDATE META │
│ - Run phase_data.py --phase-end --phase {N} │
│ - ⚠️ If validation fails: fix YAML, re-validate, iterate │
│ │
│ 9. VERIFY COMPLETENESS │
│ - All sub-tasks completed? │
│ - All issues resolved? │
│ - Output aligns with phase objective? │
│ │
│ 10. UPDATE SESSION & PROCEED │
│ - Update phases.P{N}.status = "completed" │
│ - Set phases.P{N}.completed_at = now() │
│ - Set phases.P{N}.data_file = path to YAML │
│ - Set phases.P{N}.report_file = path to MD │
│ - ✅ Only then proceed to Phase {N+1} │
│ │
└─────────────────────────────────────────────────────────────────────────┘
§3 Phase Data Contracts (YAML Files)
P1 Output: P1_project_context.yaml
# .phase_working/data/P1_project_context.yaml
schema_version: "3.0.0 (20260201a)"
phase: 1
generated_at: "ISO8601"
project_context:
project_type: "web|api|microservices|ai|llm"
project_name: "PROJECT-NAME"
tech_stack:
languages: ["Python", "TypeScript"]
frameworks: ["FastAPI", "React"]
databases: ["PostgreSQL", "Redis"]
module_inventory:
modules:
- id: M-001
name: "Authentication Module"
path: "src/auth"
type: "Authentication"
security_level: "HIGH"
files: 12
loc: 1500
entry_types: ["API", "UI"]
summary:
total_modules: 15
by_security_level: {HIGH: 5, MEDIUM: 7, LOW: 3}
entry_point_inventory:
api_entries:
- id: EP-API-001
path: "/api/v1/auth/login"
methods: ["POST"]
module: "M-001"
handler: "src/auth/handlers/login.py:45"
auth_required: false
exposure: "EXTERNAL"
ui_entries: []
system_entries: []
hidden_entries: []
summary:
total_entry_points: 72
by_type: {api: 45, ui: 12, system: 10, hidden: 5}
discovery_checklist:
rest_api: {scanned: true, count: 45, status: "COMPLETED"}
internal_api: {scanned: true, count: 8, status: "COMPLETED"}
graphql: {scanned: true, count: 0, status: "NOT_APPLICABLE"}
websocket: {scanned: true, count: 2, status: "COMPLETED"}
cron_jobs: {scanned: true, count: 3, status: "COMPLETED"}
message_queue: {scanned: true, count: 0, status: "NOT_APPLICABLE"}
webhooks: {scanned: true, count: 2, status: "COMPLETED"}
file_upload: {scanned: true, count: 3, status: "COMPLETED"}
health_endpoints: {scanned: true, count: 3, status: "COMPLETED"}
debug_endpoints: {scanned: true, count: 1, status: "COMPLETED"}
coverage: "100%"
doc_analysis: # Optional, only if docs exist
analyzed_documents: []
project_intent: {}
architecture_overview: {}
P2 Output: P2_dfd_elements.yaml
# .phase_working/data/P2_dfd_elements.yaml
schema_version: "3.0.0 (20260201a)"
phase: 2
generated_at: "ISO8601"
input_ref: "P1_project_context.yaml" # Traceability
interface_inventory:
discovery_sources:
openapi_spec: "api/openapi.yaml"
static_analysis: true
config_files: ["nginx.conf"]
interfaces:
- id: IF-001
layer: "L1"
type: "REST"
path: "/api/v1/users"
methods: ["GET", "POST", "PUT", "DELETE"]
auth_required: true
code_location: {file: "src/routes/users.py", line: 45}
input_modalities:
- type: "path_params"
params: [{name: "user_id", type: "uuid"}]
- type: "json_body"
schema_ref: "UserCreateRequest"
summary:
total_interfaces: 45
by_layer: {L1: 12, L2: 23, L3: 10}
data_flow_traces:
flows:
- id: DF-001
name: "User Creation Data Flow"
entry_interface: "IF-001"
trace:
- step: 1
component: "UserRouter.create"
action: "receive_request"
data_in: "raw_json"
data_out: "parsed_request"
- step: 2
component: "InputValidator.validate"
action: "validate_input"
security_relevant: true
sensitive_data_points:
- {step: 3, field: "password", classification: "credential"}
summary:
total_flows: 89
sensitive_flows: 23
call_flow_graph:
call_flows:
- id: CF-001
entry_interface: "IF-001"
call_chain:
- step: 1
caller: "UserRouter.create"
callee: "AuthMiddleware.verify"
security_check: true
check_type: "authentication"
security_checkpoints:
- {step: 1, type: "authentication", bypass_risk: "low"}
summary:
total_call_flows: 67
security_checkpoint_coverage: 0.72
data_store_inventory:
data_stores:
- id: DS-001
name: "Primary PostgreSQL"
type: "relational_database"
technology: "PostgreSQL"
sensitive_data:
- table: "users"
fields:
- {name: "password_hash", classification: "credential"}
- {name: "email", classification: "PII"}
security_config:
encryption_at_rest: true
encryption_in_transit: true
summary:
total_data_stores: 5
encryption_coverage: 0.83
dfd_elements:
external_interactors:
- id: EI-001
name: "Web User"
type: "Human"
trust_level: "untrusted"
processes:
- id: P-001
name: "API Gateway"
layer: "L1"
maps_to_module: "M-001"
data_stores:
- id: DS-001
name: "User Database"
type: "PostgreSQL"
sensitivity: "HIGH"
data_flows:
- id: DF-001
from: "EI-001"
to: "P-001"
data: "User Request"
encrypted: true
l1_coverage:
total_interfaces: 45
analyzed: 45
coverage_percentage: 100
interface_analysis:
IF-001:
analyzed: true
data_flow_traced: true
call_flow_mapped: true
validation_results:
interface_coverage: 1.0
data_flow_coverage: 1.0
recommendation: "PROCEED_TO_PHASE_3"
P3 Output: P3_boundary_context.yaml
# .phase_working/data/P3_boundary_context.yaml
schema_version: "3.0.0 (20260201a)"
phase: 3
generated_at: "ISO8601"
input_ref: "P2_dfd_elements.yaml"
boundary_context:
boundaries:
- id: TB-001
name: "External Trust Boundary"
type: "network"
scope: "internet-facing"
components_inside: ["P-001"]
components_outside: ["EI-001", "EI-002"]
interfaces:
- id: TBI-001
boundary: "TB-001"
direction: "inbound"
from_zone: "external"
to_zone: "dmz"
data_flows: ["DF-001", "DF-002"]
protection: ["authentication", "rate_limiting"]
data_nodes:
- id: DN-001
data_store: "DS-001"
sensitivity: "HIGH"
data_types: ["credential", "PII"]
boundary_location: "internal"
cross_boundary_flows:
- id: CBF-001
flow: "DF-001"
crosses: ["TB-001"]
risk_level: "HIGH"
P4 Output: P4_security_gaps.yaml
# .phase_working/data/P4_security_gaps.yaml
schema_version: "3.0.0 (20260201a)"
phase: 4
generated_at: "ISO8601"
input_ref: "P3_boundary_context.yaml"
security_gaps:
gaps:
- id: GAP-001
domain: "AUTHN"
severity: "HIGH"
description: "No MFA for admin accounts"
affected_components: ["P-002"]
recommendation: "Implement TOTP-based MFA"
design_matrix:
AUTHN:
score: 0.7
controls_present: ["password_auth", "jwt_tokens"]
controls_missing: ["mfa", "passwordless"]
AUTHZ:
score: 0.8
controls_present: ["rbac"]
controls_missing: ["abac", "policy_engine"]
# ... 16 domains total
findings:
- id: F-P4-001
domain: "CRYPTO"
severity: "MEDIUM"
observation: "Using SHA-256 for password hashing"
expected: "bcrypt/argon2"
P5 Output: P5_threat_inventory.yaml
# .phase_working/data/P5_threat_inventory.yaml
schema_version: "3.0.0 (20260201a)"
phase: 5
generated_at: "ISO8601"
input_ref: "P4_security_gaps.yaml"
threat_inventory:
threats:
- id: T-S-P001-001
stride_category: "S"
element_type: "process"
element_id: "P-001"
title: "API Gateway Spoofing"
description: "Attacker impersonates legitimate API client"
cwe_refs: ["CWE-287", "CWE-290"]
capec_refs: ["CAPEC-151"]
attack_refs: ["T1078"]
affected_flows: ["DF-001"]
related_gaps: ["GAP-001"]
initial_risk: "HIGH"
# ... more threats
summary:
total: 45
by_stride: {S: 8, T: 12, R: 5, I: 10, D: 6, E: 4}
by_element_type: {process: 25, datastore: 12, dataflow: 8}
by_risk: {CRITICAL: 3, HIGH: 15, MEDIUM: 20, LOW: 7}
P6 Output: P6_validated_risks.yaml
# .phase_working/data/P6_validated_risks.yaml
schema_version: "3.0.0 (20260201a)"
phase: 6
generated_at: "ISO8601"
input_ref: "P5_threat_inventory.yaml"
risk_summary:
total_identified: 45 # From P5
total_verified: 15 # Confirmed exploitable
total_theoretical: 20 # Possible but unconfirmed
total_pending: 5 # Need more investigation
total_excluded: 5 # False positives or mitigated
# MUST BALANCE: 15 + 20 + 5 + 5 = 45
risk_details:
- id: VR-001
threat_refs: ["T-S-P001-001"]
validation_status: "verified"
validation_method: "code_review"
evidence: "No session validation in auth middleware"
cvss_score: 8.1
risk_level: "HIGH"
poc_details:
- id: POC-001
risk_ref: "VR-001"
poc_type: "exploit_code"
description: "JWT token forgery using weak secret"
steps:
- "Intercept valid JWT token"
- "Decode and modify claims"
- "Re-sign with guessed secret"
success_criteria: "Gain admin access"
estimated_effort: "LOW"
attack_paths:
- id: AP-001
name: "Privilege Escalation via Token Forgery"
entry_point: "IF-001"
target: "DS-001"
steps:
- {step: 1, action: "Obtain valid user JWT", component: "P-002"}
- {step: 2, action: "Forge admin claims", component: "external"}
- {step: 3, action: "Access admin endpoints", component: "P-003"}
feasibility: "HIGH"
impact: "CRITICAL"
attack_chains:
- id: AC-001
name: "Full Database Compromise Chain"
paths: ["AP-001", "AP-002"]
total_steps: 5
complexity: "MEDIUM"
P7 Output: P7_mitigation_plan.yaml
# .phase_working/data/P7_mitigation_plan.yaml
schema_version: "3.0.0 (20260201a)"
phase: 7
generated_at: "ISO8601"
input_ref: "P6_validated_risks.yaml"
mitigation_plan:
mitigations:
- id: MIT-001
risk_refs: ["VR-001"]
title: "Implement Strong JWT Secret Management"
priority: "P0"
effort: "MEDIUM"
control_type: "preventive"
implementation:
description: "Use cryptographically strong secrets with rotation"
code_changes:
- file: "src/auth/jwt.py"
change_type: "modify"
description: "Add secret rotation logic"
config_changes:
- file: "config/auth.yaml"
change_type: "add"
description: "Add JWT_SECRET_ROTATION_DAYS"
verification:
test_cases: ["TC-001", "TC-002"]
acceptance_criteria: "JWT secrets rotate every 30 days"
roadmap:
immediate: # P0: Fix now
- "MIT-001"
- "MIT-002"
short_term: # P1: Within 7 days
- "MIT-003"
medium_term: # P2: Within 30 days
- "MIT-004"
- "MIT-005"
long_term: # P3: Backlog
- "MIT-006"
P8 Output: P8_report_manifest.yaml
# .phase_working/data/P8_report_manifest.yaml
schema_version: "3.0.0 (20260201a)"
phase: 8
generated_at: "ISO8601"
input_ref: "P7_mitigation_plan.yaml"
report_manifest:
main_report: "RISK-ASSESSMENT-REPORT.md"
generated_reports:
- name: "RISK-INVENTORY.md"
source_data: "P6_validated_risks.yaml"
status: "generated"
- name: "MITIGATION-MEASURES.md"
source_data: "P7_mitigation_plan.yaml"
status: "generated"
- name: "PENETRATION-TEST-PLAN.md"
source_data: ["P6_validated_risks.yaml", "P7_mitigation_plan.yaml"]
status: "generated"
- name: "ARCHITECTURE-ANALYSIS.md"
source_data: ["P1_project_context.yaml", "P2_dfd_elements.yaml", "P3_boundary_context.yaml"]
status: "generated"
- name: "DFD-DIAGRAM.md"
source_data: "P2_dfd_elements.yaml"
status: "generated"
- name: "COMPLIANCE-REPORT.md"
source_data: ["P4_security_gaps.yaml", "P7_mitigation_plan.yaml"]
status: "generated"
- name: "ATTACK-PATH-VALIDATION.md"
source_data: "P6_validated_risks.yaml"
status: "generated"
phase_reports_published:
- "P1-PROJECT-UNDERSTANDING.md"
- "P2-DFD-ANALYSIS.md"
- "P3-TRUST-BOUNDARY.md"
- "P4-SECURITY-DESIGN-REVIEW.md"
- "P5-STRIDE-THREATS.md"
- "P6-RISK-VALIDATION.md"
statistics:
total_threats: 45
verified_risks: 15
mitigations_planned: 25
compliance_score: 0.78
§4 Validation Gates
Exit Codes
| Code | Meaning | Action |
|---|---|---|
| 0 | Pass | Proceed to next phase |
| 1 | Missing data | Fix YAML and revalidate |
| 2 | Schema validation failed | Fix structure and revalidate |
Phase-Specific Validation
| Phase | Data File Required | Validation Criteria |
|---|---|---|
| 1 | P1_project_context.yaml | module_inventory + entry_point_inventory + discovery_checklist |
| 2 | P2_dfd_elements.yaml | l1_coverage.coverage_percentage == 100 |
| 3 | P3_boundary_context.yaml | boundaries[] non-empty |
| 4 | P4_security_gaps.yaml | design_matrix with 16 domains |
| 5 | P5_threat_inventory.yaml | threat_inventory.summary.total > 0 |
| 6 | P6_validated_risks.yaml | Count conservation balanced |
| 7 | P7_mitigation_plan.yaml | Every VR-xxx has MIT-xxx |
| 8 | P8_report_manifest.yaml | All 8 reports generated |
Count Conservation
P5.threat_inventory.summary.total ==
P6.risk_summary.total_verified +
P6.risk_summary.total_theoretical +
P6.risk_summary.total_pending +
P6.risk_summary.total_excluded
§5 STRIDE per Element Matrix
| Element Type | S | T | R | I | D | E |
|---|---|---|---|---|---|---|
| Process | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Data Store | ✓ | ✓ | ✓ | ✓ | ||
| Data Flow | ✓ | ✓ | ✓ | |||
| External (source) | ✓ | ✓ |
§6 Error Recovery
Validation Failure
- Read error message from validation output
- Identify missing/invalid fields in YAML
- Fix YAML data file
- Re-run validation
- Only then update report
Session Interruption
- Check
.phase_working/{SESSION_ID}/_session_meta.yaml - Find last phase with
status: "completed" - Load that phase's YAML data file from
data/subdirectory - Resume from next phase
SESSION_ID 格式:
{PROJECT_NAME}_{YYYYMMDD_HHMMSS}示例:OPEN-WEBUI_20260130_143022
§7 Final Report Output
Directory Structure
Risk_Assessment_Report/
│
│ ┌─ 必需报告 (4份) ────────────────────────────────────────────┐
├── {PROJECT}-RISK-ASSESSMENT-REPORT.md ← Main report
├── {PROJECT}-RISK-INVENTORY.md ← From P6 YAML
├── {PROJECT}-MITIGATION-MEASURES.md ← From P7 YAML
├── {PROJECT}-PENETRATION-TEST-PLAN.md ← From P6 YAML
│ └─────────────────────────────────────────────────────────────┘
│
│ ┌─ 扩展报告 (可选) ───────────────────────────────────────────┐
├── {PROJECT}-ARCHITECTURE-ANALYSIS.md ← From P1-P3 YAML
├── {PROJECT}-DFD-DIAGRAM.md ← From P2 YAML
├── {PROJECT}-COMPLIANCE-REPORT.md ← From P4+P7 YAML
├── {PROJECT}-ATTACK-PATH-VALIDATION.md ← From P6 YAML
│ └─────────────────────────────────────────────────────────────┘
│
│ ┌─ 阶段过程文档 (7份) ────────────────────────────────────────┐
├── P1-PROJECT-UNDERSTANDING.md ← Human report
├── P2-DFD-ANALYSIS.md
├── P3-TRUST-BOUNDARY.md
├── P4-SECURITY-DESIGN-REVIEW.md
├── P5-STRIDE-THREATS.md
├── P6-RISK-VALIDATION.md
├── P7-MITIGATION-PLAN.md
│ └─────────────────────────────────────────────────────────────┘
│
└── .phase_working/
│
├── _sessions_index.yaml ← 多 session 索引 (可选)
│
└── {SESSION_ID}/ ← Session 隔离目录
│ 格式: {PROJECT}_{YYYYMMDD_HHMMSS}
│ 示例: OPEN-WEBUI_20260130_143022
│
├── _session_meta.yaml ← Session 状态元数据
│
├── data/ ← STRUCTURED DATA (YAML)
│ ├── P1_project_context.yaml
│ ├── P2_dfd_elements.yaml
│ ├── P3_boundary_context.yaml
│ ├── P4_security_gaps.yaml
│ ├── P5_threat_inventory.yaml
│ ├── P6_validated_risks.yaml
│ ├── P7_mitigation_plan.yaml
│ └── P8_report_manifest.yaml
│
└── reports/ ← WORKING REPORTS (MD)
└── (phase reports during execution)
Session ID 生成规则
# Session ID 格式
session_id: "{PROJECT_NAME}_{YYYYMMDD_HHMMSS}"
# 示例
session_id: "OPEN-WEBUI_20260130_143022"
# 路径构造
session_dir: "Risk_Assessment_Report/.phase_working/{SESSION_ID}"
data_dir: "{session_dir}/data"
reports_dir: "{session_dir}/reports"
meta_file: "{session_dir}/_session_meta.yaml"
多 Session 支持
同一项目可以有多个评估 session：
.phase_working/
├── _sessions_index.yaml ← 可选：列出所有 sessions
├── OPEN-WEBUI_20260128_091500/ ← 历史 session
├── OPEN-WEBUI_20260129_142030/ ← 历史 session
└── OPEN-WEBUI_20260130_143022/ ← 当前 session (active)
End of WORKFLOW.md (~400 lines, ~5K tokens)