跳到主要内容
知仓学习社ZHICANG

gsd-tools

Central utility skill for GSD operations. Provides config parsing, slug generation, timestamps, path operations, and orchestrates calls to other spe…

执行命令严重 1 · 高危 0a5c-ai/babysitter

它会碰到什么

扫了多少2 个文本文件,10 KB
它会碰到什么执行命令
命中总数1 处
命中统计严重 1 · 高 0 · 中 0 · 低 0
逐条看命中(1 条严重或高危)
  • 严重 SKILL.md:3perm-wildcard
    allowed-tools: Bash(*) Read Write Edit Glob Grep

这一栏是扫描器报的事实,不是结论。命中多不等于有毒(安全工具、规则库、示例脚本本来就会包含危险写法),命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系,需要你自己看。

技能内容

  • Load and manage .planning/config.json configuration
  • Generate deterministic slugs from descriptions (kebab-case, deduplication)
  • Format timestamps for GSD artifact headers and frontmatter
  • Resolve paths within the .planning/ directory structure
  • Parse and normalize phase numbers (integer and decimal)
  • Manage sequential quick task numbering in .planning/quick/
  • Initialize and validate .planning/ directory structure
  • Detect project state (new project, existing project, mid-phase, etc.)

Capabilities

1. Configuration Loading

Load and parse .planning/config.json with defaults:

{
  "profile": "balanced",
  "autoCommit": true,
  "autoVerify": true,
  "planCheckerEnabled": true,
  "maxPlanRevisions": 2,
  "waveParallelization": true,
  "contextMonitor": true,
  "contextWarningThreshold": 70,
  "contextCriticalThreshold": 85,
  "summaryVariant": "standard",
  "tddEnabled": false,
  "questioningDepth": "adaptive"
}

Read config:

cat .planning/config.json

If config does not exist, create with defaults. Merge user overrides with defaults (user values take precedence).

2. Slug Generation

Generate kebab-case slugs from descriptions:

Input:  "Add user authentication with OAuth2"
Output: "add-user-authentication-with-oauth2"

Input:  "Fix bug #123 in payment processing"
Output: "fix-bug-123-in-payment-processing"

Rules:

  • Lowercase all characters
  • Replace spaces and special characters with hyphens
  • Remove consecutive hyphens
  • Trim leading/trailing hyphens
  • Maximum 60 characters (truncate at word boundary)
  • Check for duplicates in target directory, append -2, -3 if needed

3. Timestamp Formatting

Format timestamps for GSD artifacts:

Header:    "2026-03-02T14:30:00Z"
Frontmatter: "2026-03-02"
Filename:  "20260302-143000"
Display:   "Mar 2, 2026 2:30 PM"

4. Path Operations

Resolve paths within .planning/ structure:

Phase directory:    .planning/phase-{N}/
Plan file:          .planning/phase-{N}/PLAN-{M}.md
Summary file:       .planning/phase-{N}/SUMMARY.md
Quick task:         .planning/quick/{NNN}-{slug}/
Debug session:      .planning/debug/{slug}.md
Codebase docs:      .planning/codebase/
Research docs:      .planning/research/
Milestone archive:  milestones/v{X}.{Y}/

5. Phase Number Parsing

Parse integer and decimal phase numbers:

"72"    -> { major: 72, minor: null, display: "72" }
"72.1"  -> { major: 72, minor: 1, display: "72.1" }
"72.2"  -> { major: 72, minor: 2, display: "72.2" }

Decimal phases are inserted phases (e.g., urgent work between phase 72 and 73).

6. Quick Task Numbering

Sequential numbering for quick tasks:

.planning/quick/001-fix-login-bug/
.planning/quick/002-add-rate-limiting/
.planning/quick/003-update-readme/

Scan existing directories, find highest number, increment by 1. Zero-pad to 3 digits.

7. Planning Directory Initialization

Ensure .planning/ directory exists with required structure:

.planning/
  config.json
  PROJECT.md       (if new-project has run)
  REQUIREMENTS.md  (if new-project has run)
  ROADMAP.md       (if new-project has run)
  STATE.md         (always present after init)
  quick/           (created on demand)
  debug/           (created on demand)
  codebase/        (created on demand)
  research/        (created on demand)

8. Project State Detection

Detect current project state for routing:

const states = {
  NO_PROJECT: "No .planning/ directory or no PROJECT.md",
  INITIALIZED: ".planning/ exists with PROJECT.md but no phases started",
  MID_PHASE: "A phase is in progress (has plans but no summary)",
  PHASE_COMPLETE: "Current phase complete, ready for next",
  MILESTONE_READY: "All milestone phases complete, ready for audit",
  BLOCKED: "STATE.md has active blockers"
};

Tool Use Instructions

Reading Configuration

  1. Use Read to load .planning/config.json
  2. Parse JSON and merge with defaults
  3. Return merged configuration object

Generating Slugs

  1. Accept description string as input
  2. Apply slug transformation rules
  3. Use Glob to check for duplicates in target directory
  4. Append deduplication suffix if needed
  5. Return final slug

Initializing Planning Directory

  1. Use Bash to create directory structure with mkdir -p
  2. Use Write to create config.json with defaults if missing
  3. Use Read + Glob to detect existing state
  4. Return initialization status

Process Integration

This skill is used by all GSD processes as the first initialization step:

  • new-project.js - Initialize .planning/, create config
  • plan-phase.js - Resolve phase paths, load config for planner settings
  • execute-phase.js - Load config for executor settings, resolve task paths
  • quick.js - Generate quick task number and slug, resolve quick task path
  • debug.js - Generate debug session slug, resolve debug path
  • progress.js - Detect project state, load config
  • complete-milestone.js - Resolve milestone archive path
  • add-tests.js - Resolve phase paths for test generation
  • research-phase.js - Resolve research output paths

Output Format

{
  "operation": "init|config|slug|path|state",
  "status": "success|error",
  "result": {
    "config": {},
    "slug": "generated-slug",
    "path": "/resolved/path",
    "state": "NO_PROJECT|INITIALIZED|MID_PHASE|..."
  },
  "metadata": {
    "planningDir": ".planning/",
    "configPath": ".planning/config.json",
    "timestamp": "2026-03-02T14:30:00Z"
  }
}

Configuration

| Setting | Default | Description |

|---------|---------|-------------|

| profile | balanced | Model profile (quality/balanced/budget) |

| autoCommit | true | Auto-commit after each task |

| autoVerify | true | Auto-verify after phase execution |

| planCheckerEnabled | true | Run plan checker before execution |

| maxPlanRevisions | 2 | Max plan revision cycles |

| waveParallelization | true | Enable wave-based parallel execution |

| contextMonitor | true | Enable context window monitoring |

| summaryVariant | standard | Summary template variant |

Error Handling

| Error | Cause | Resolution |

|-------|-------|------------|

| Config parse error | Malformed config.json | Reset to defaults, warn user |

| Planning dir not writable | Permission issue | Check filesystem permissions |

| Slug collision limit | 10+ duplicates with same slug | Use timestamp suffix instead |

| Phase number invalid | Non-numeric phase argument | Show valid format examples |

| State detection failure | Corrupt STATE.md | Suggest health command to repair |

Constraints

  • Never modify files outside .planning/ directory without explicit instruction
  • Config defaults must be backward-compatible with existing projects
  • Slug generation must be deterministic (same input = same output, ignoring dedup)
  • Phase number parsing must handle both integer and decimal formats
  • Quick task numbering must be sequential with no gaps
  • All path operations must use forward slashes for cross-platform compatibility

想直接用这个技能?

本站把开放许可(MIT / Apache 等)的技能按仓库打包整理到网盘,点一下转存到你自己的网盘,不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接,不打包。