AgentSkillsCN

sprint

基于技术规范的自动化冲刺规划与执行（提示生成、依赖关系规划、有状态执行）

SKILL.md
--- frontmatter
name: sprint
description: Automated sprint planning and execution from technical specifications (prompt generation, dependency planning, stateful execution)
allowed-tools:
  - Bash(git:*)
  - Bash(jq:*)
  - Bash(npx cclimits:*)
  - Bash(python3:*)
  - Bash(find:*)
  - Bash(ls:*)
  - Bash(cat:*)
  - Bash(mkdir:*)
  - Bash(realpath:*)
  - Read
  - Write
  - Edit
  - Glob
  - Grep

Sprint Planning + Execution

This skill turns a technical specification into:

  • •Prompt files (ready for /run-prompt)
  • •A dependency-aware execution plan (sprint-plan.md)
  • •A persistent state file (.sprint-state.json) for long-running execution

CLI

Main command

bash
# From a spec (generates new prompts)
python3 skills/sprint/scripts/sprint.py <spec-file-or-text> [options]

# From existing prompts (no spec argument)
python3 skills/sprint/scripts/sprint.py --from-existing [options]

Options:

  • •--output-dir DIR (default: ./prompts/) Where to write generated prompt files
  • •--plan-file FILE (default: ./sprint-plan.md) Where to write the plan markdown
  • •--dry-run Generate plan without creating prompt files or state
  • •--from-existing Analyze existing prompts in --output-dir instead of generating from a spec
  • •--prompts LIST Include only specific prompts (e.g., 001-005,010)
  • •--folder PATH Only include prompts from this subfolder of --output-dir (e.g., providers/)
  • •--exclude LIST Exclude specific prompts (e.g., 003,007)
  • •--auto-execute Execute phases immediately, updating .sprint-state.json
  • •--models LIST Comma-separated models (default: claude,codex,gemini)
  • •--max-parallel N Max concurrent prompts per phase (default: 5)
  • •--worktree Use worktree isolation when auto-executing
  • •--loop Use verification loops when auto-executing
  • •--max-iterations N (default: 3) Verification loop max iterations
  • •--completion-marker TEXT (default: VERIFICATION_COMPLETE) Loop completion marker
  • •--state-file FILE (default: .sprint-state.json) State file path
  • •--json Print JSON output (plan + state summary)

Sub-commands

bash
python3 skills/sprint/scripts/sprint.py status
python3 skills/sprint/scripts/sprint.py add "Implement caching layer"
python3 skills/sprint/scripts/sprint.py remove 005
python3 skills/sprint/scripts/sprint.py replan
python3 skills/sprint/scripts/sprint.py pause
python3 skills/sprint/scripts/sprint.py resume
python3 skills/sprint/scripts/sprint.py cancel --yes
python3 skills/sprint/scripts/sprint.py history

Notes:

  • •Sub-commands operate on .sprint-state.json in the current directory unless --state-file is provided.
  • •cancel is destructive for sprint-created worktrees (it removes worktree directories/branches recorded in state).

Examples

bash
# Analyze all prompts in prompts/
python3 skills/sprint/scripts/sprint.py --from-existing --dry-run

# Only specific prompts
python3 skills/sprint/scripts/sprint.py --from-existing --prompts 001-005,010 --dry-run

# Only prompts in a subfolder
python3 skills/sprint/scripts/sprint.py --from-existing --folder providers/ --dry-run

# Exclude certain prompts
python3 skills/sprint/scripts/sprint.py --from-existing --exclude 003,007 --dry-run

# Combine with execution options
python3 skills/sprint/scripts/sprint.py --from-existing --worktree --loop --auto-execute