roadmap

majiayu000 的头像作者 majiayu000

查看并更新开发路线图。在每次会议开始时使用,以了解当前进度和下一步的优先事项。

📖 中文导读:查看并更新开发路线图。在每次会议开始时使用,以了解当前进度和下一步的优先事项。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: roadmap description: View and update the development roadmap. Use at the start of each session to understand current progress and next priorities.

GATE 4: PLANNING — PLANNER PROTOCOL

Agent: PLANNER Gate: 4 of 6 Prerequisite: Gate 3 (Test Design) COMPLETE Output: docs/planning/ROADMAP.md with traced tasks


GATE 4 ENTRY CHECKLIST

Before proceeding, verify:

  • .fortress/gates/GATE_3_TEST_DESIGN.md exists
  • All test stubs exist in tests/
  • TEST_MATRIX.md is complete
  • pytest --collect-only succeeds

If any checkbox fails: STOP. Complete Gate 3 first.


PLANNER PROTOCOL

Core Principle: TASKS TRACE TO SPECS

❌ FORBIDDEN: "Implement feature X"
✅ REQUIRED: "Implement S001 - Package validation"
             TRACES: S001, INV001, INV002
             TESTS: T001.1-T001.10

Every task MUST link to:

  1. SPEC_ID(s) it implements
  2. INV_ID(s) it enforces
  3. TEST_ID(s) it must pass

Step 1: Define Task Size Limits

Rule Limit Reason
Max task size 8 hours Smaller = better estimates
Estimation multiplier 2.5x Always over-estimate
Buffer per week 20% Unexpected issues

Step 2: Create Task Breakdown

For each SPEC_ID, create implementation tasks:

## TASK: W1.1 — Implement Core Types

### Traces
- SPEC: S001, S002
- INVARIANTS: INV001, INV002
- TESTS: T001.1-T001.5

### Definition
Implement the core data types for PackageRisk, Signals, and Recommendation.

### Acceptance Criteria
- [ ] T001.1 passes (valid package structure)
- [ ] T001.2 passes (empty input handling)
- [ ] INV001 enforced (risk_score bounds)
- [ ] INV002 enforced (signals not None)
- [ ] mypy passes with no errors
- [ ] 100% coverage on new code

### Estimated Effort
- Optimistic: 2 hours
- Likely: 4 hours
- Pessimistic: 8 hours
- **Planned**: 4 hours × 2.5 = 10 hours

### Dependencies
- None (first task)

### Pre-Conditions
- Test stubs T001.1-T001.5 exist (GATE 3)
- Types defined in architecture (GATE 1)

### Post-Conditions
- Tests T001.1-T001.5 pass (not skipped)
- src/phantom_guard/core/types.py exists
- Code has IMPLEMENTS: S001 comments

Step 3: Create Week Plans

Group tasks into weeks:

# docs/planning/ROADMAP.md

## Week 1: Core Types & Detection Engine

### Goals
- Implement S001-S005 (Core detection)
- All T001.* tests passing
- 90% coverage on core module

### Tasks

| Task | SPEC | Hours | Status |
|:-----|:-----|:------|:-------|
| W1.1 | S001 | 10 | PENDING |
| W1.2 | S002 | 8 | PENDING |
| W1.3 | S003 | 6 | PENDING |
| W1.4 | S004 | 8 | PENDING |
| W1.5 | S005 | 8 | PENDING |
| **Buffer** | - | 10 | - |
| **Total** | - | 50 | - |

### Exit Criteria
- [ ] All W1.* tasks complete
- [ ] T001-T005 tests passing
- [ ] Coverage ≥90% on core/
- [ ] Hostile review passed

---

## Week 2: Registry Clients

### Goals
- Implement S030-S055 (All registry clients)
- All T030-T055 tests passing
- Integration tests with live APIs

### Tasks

| Task | SPEC | Hours | Status |
|:-----|:-----|:------|:-------|
| W2.1 | S030-S035 | 12 | PENDING |
| W2.2 | S040-S045 | 12 | PENDING |
| W2.3 | S050-S055 | 12 | PENDING |
| W2.4 | Cache integration | 6 | PENDING |
| **Buffer** | - | 8 | - |
| **Total** | - | 50 | - |

### Exit Criteria
- [ ] All W2.* tasks complete
- [ ] T030-T055 tests passing
- [ ] Live API tests work
- [ ] Hostile review passed

Step 4: Create Dependency Graph

## DEPENDENCY GRAPH

W1.1 (Core Types) │ ├── W1.2 (Risk Scoring) ──┐ │ │ ├── W1.3 (Signal Detection)──┤ │ │ └── W1.4 (Pattern Matching)──┘ │ ▼ W1.5 (Detector Integration) │ ┌──────────────────┼──────────────────┐ │ │ │ ▼ ▼ ▼ W2.1 (PyPI) W2.2 (npm) W2.3 (crates) │ │ │ └──────────────────┼──────────────────┘ │ ▼ W2.4 (Cache Integration) │ ▼ W3.1 (CLI Interface)


### Critical Path
W1.1 → W1.5 → W2.1 → W2.4 → W3.1 → W4.1 (Release)

Step 5: Risk Assessment

## RISK ASSESSMENT

| Task | Risk Level | Risk Factor | Mitigation |
|:-----|:-----------|:------------|:-----------|
| W1.1 | LOW | Well-defined types | N/A |
| W1.2 | MEDIUM | Algorithm complexity | Property tests |
| W2.1 | HIGH | External API | Mocking, retries |
| W2.2 | HIGH | External API | Mocking, retries |
| W3.1 | LOW | Standard CLI patterns | typer framework |

### Contingency Plans

**If PyPI API changes:**
1. Switch to cached-only mode
2. Document regression
3. Implement new parser

**If rate limiting hit:**
1. Implement backoff
2. Add queue mechanism
3. Consider API key

ROADMAP DOCUMENT TEMPLATE

# Phantom Guard — Development Roadmap

> **Version**: 0.1.0
> **Created**: YYYY-MM-DD
> **Last Updated**: YYYY-MM-DD
> **Status**: ACTIVE

---

## Overview

### MVP Target
- Date: YYYY-MM-DD
- Scope: Core detection + CLI + 3 registries

### Total Effort
- Planned: X weeks
- Buffer: Y% contingency

---

## Phase 1: Core Engine (Weeks 1-2)

[Detailed week plans with tasks]

## Phase 2: Integrations (Weeks 3-4)

[Detailed week plans with tasks]

## Phase 3: CLI & Hooks (Weeks 5-6)

[Detailed week plans with tasks]

## Phase 4: Polish & Release (Weeks 7-8)

[Release preparation tasks]

---

## Task Registry

| Task ID | SPEC | Description | Hours | Status | Week |
|:--------|:-----|:------------|:------|:-------|:-----|
| W1.1 | S001 | Core types | 10 | PENDING | 1 |
| W1.2 | S002 | Risk scoring | 8 | PENDING | 1 |
| ... | ... | ... | ... | ... | ... |

---

## Progress Tracking

### Week 1
- [ ] W1.1 Complete
- [ ] W1.2 Complete
- [ ] Week 1 hostile review

### Week 2
...

---

## Trace Matrix

| Task | SPEC_IDs | INV_IDs | TEST_IDs |
|:-----|:---------|:--------|:---------|
| W1.1 | S001 | INV001, INV002 | T001.1-T001.5 |
| ... | ... | ... | ... |

GATE 4 EXIT CHECKLIST

Before Gate 4 is complete:

  • docs/planning/ROADMAP.md exists
  • Every task has SPEC trace
  • Every task has TEST trace
  • Every task < 8 hours
  • 20% buffer per week
  • Dependency graph defined
  • Risk assessment complete
  • HOSTILE_PLANNER review passed

If any checkbox fails: DO NOT PROCEED TO IMPLEMENTATION.


HOSTILE_PLANNER REVIEW

After completing roadmap:

/hostile-review planning

The reviewer checks:

  • Are estimates realistic?
  • Are dependencies correct?
  • Are traces complete?
  • Are risks identified?
  • Is critical path clear?

RECORDING GATE COMPLETION

# .fortress/gates/GATE_4_PLANNING.md

## Gate 4: Planning — COMPLETE

**Date**: YYYY-MM-DD
**Approver**: HOSTILE_PLANNER
**Output**: docs/planning/ROADMAP.md

### Summary
- X tasks defined
- Y weeks planned
- Z hours total (with buffer)

### Schedule
| Phase | Weeks | Hours |
|:------|:------|:------|
| Core | 1-2 | X |
| Integrations | 3-4 | Y |
| CLI | 5-6 | Z |
| Release | 7-8 | W |

### Next Step
Begin implementation with /implement W1.1

IMPLEMENTATION WORKFLOW

After Gate 4, use /implement for each task:

/implement W1.1

This triggers:

  1. Load task definition
  2. Verify pre-conditions
  3. TDD implementation cycle
  4. Verify post-conditions
  5. Mark task complete

PROTOCOL VIOLATIONS

Violation Response
Task without SPEC trace Add trace
Task > 8 hours Split task
No buffer in week Add 20% buffer
Skipped HOSTILE review Run review
Starting implementation without Gate 4 BLOCKED

Gate 4 is about PLANNING what to build. Gate 5 is about VALIDATING what was built.