AgentSkillsCN

streaks

当通过Streaks CLI（`st`）管理Streaks应用的习惯时使用：探索行动、关联快捷方式、执行习惯/任务动作、配置输出（Agent NDJSON、JSON UTI）、排查安装/诊断/探索问题，或解读退出码与JSON封装，以实现自动化操作。

SKILL.md
--- frontmatter
name: streaks
description: "Use when managing Streaks app habits via the Streaks CLI (`st`): discover actions, link Shortcuts, run habit/task actions, configure output (agent NDJSON, JSON UTI), troubleshoot install/doctor/discover, or interpret exit codes and JSON envelopes for automation."

Streaks CLI

Overview

Use this skill to manage Streaks app habits via the st CLI with Shortcuts-based actions and the streaks:// URL scheme. The st CLI must be installed to use this skill.

Install the CLI with Homebrew (brew tap vburojevic/tap then brew install streaks-cli) or from source (make build).

Quick start

  • •Run st discover and st doctor to validate app + Shortcuts readiness.
  • •Run st open to confirm Streaks launches.
  • •Run st help or st help <command> to resolve usage questions.

Core workflow

  1. •Discover capabilities: st discover (use --agent for NDJSON).
  2. •Verify setup: st doctor.
  3. •Ensure shortcuts exist:
    • •If missing, create them in Shortcuts or import wrapper .shortcut files with st install --import.
  4. •Map custom shortcut names with st link <action-id> --shortcut "Name".
  5. •Run actions: st <action> [--task "Habit"].
  6. •When automating, enable agent mode and parse the last NDJSON line.

Running actions

  • •Use st actions list and st actions describe <action> to inspect available actions and required parameters.
  • •Use --task for habit/task-based actions.
  • •Use --input or --stdin to pass raw JSON input.
  • •Use --dry-run to verify shortcut name + payload before execution.
  • •Use --trace <file> to append JSONL trace records.

Output and automation

  • •Default output is human-oriented for meta commands and raw shortcut output for actions.
  • •Use --agent or STREAKS_CLI_AGENT=1 for NDJSON envelopes and stable errors.
  • •Use --shortcuts-output public.json when you need JSON data from Shortcuts.
  • •Use --no-output when only exit codes matter.

Troubleshooting

  • •Use st doctor for readiness checks.
  • •Use st discover --markdown for a human-readable capability report.
  • •If an action fails, check stderr JSON in agent mode and consult exit codes.

References

  • •Use references/commands.md for the full command/flag list and env vars.
  • •Use references/agent.md for NDJSON conventions and integration patterns.
  • •Use references/schema.md for exit codes and envelope schemas.
  • •Use references/troubleshooting.md for common errors and fixes.