Skip to content
Agent Orchestration
Skill

/always-on-guidance

Always-on rule-oriented guidance for claude-plugin agents. Use to align behavior, tool usage, and model-specific defaults while avoiding deprecated bd/cass references. Related skills: swarm-coordination, testing-patterns.

From plugin
swarm-tools
7298 skills3 agents6 commands1 MCP
Install
$ npx -y skills add joelhooks/swarm-tools --skill always-on-guidance --agent claude-code

How it fires

How this skill gets triggered: by you, by Claude, or both.

  • Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
  • You can call itInvoke it directly when you want it.
  • Slash command/always-on-guidance

Context preview

The summary Claude sees to decide when to auto-load this skill.

Always-on rule-oriented guidance for claude-plugin agents. Use to align behavior, tool usage, and model-specific defaults while avoiding deprecated bd/cass references. Related skills: swarm-coordination, testing-patterns.

SKILL.md

always-on-guidance.SKILL.md
name: always-on-guidance
description: |
  Always-on rule-oriented guidance for claude-plugin agents. Use to align behavior,
  tool usage, and model-specific defaults while avoiding deprecated bd/cass references.
  Related skills: swarm-coordination, testing-patterns.

Always-On Guidance

Global Rules

  • Follow instruction priority: system → developer → user → AGENTS.
  • Use swarm plugin tools (`hive_*`, `swarm_*`, `swarmmail_*`, `hivemind_*`); avoid deprecated `bd`/`cass` references.
  • Stay within assigned files; reserve before edits with `ttl_seconds`; release reservations on done; finish swarm work with `swarm_complete`.
  • Use `TaskCreate`/`TaskUpdate` for visible progress in Claude Code UI alongside `hive_*` for git-backed persistence.
  • When `CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS` is enabled, prefer `TeammateTool` for real-time coordination and `swarmmail_*` for persistence.
  • `swarmmail_release_all` is coordinator-only for stale/orphaned reservations.
  • Keep outputs concise and action-oriented.

Model Defaults

Use model aliases (`inherit`, `opus`, `sonnet`, `haiku`) instead of version numbers.

Opus

  • Allow brief rationale (1–2 sentences) for decisions.
  • Use sections when work has multiple phases.
  • Suggest alternatives only when risk is high, then choose one.
  • Stay compact; avoid long exposition.

Sonnet/Haiku

  • Prefer strict checklists and short imperatives.
  • Ask a single clarifying question if blocked; otherwise proceed.
  • Avoid speculative reasoning; state decisions plainly.
  • Keep outputs minimal and non-narrative.

Testing Discipline

  • Use red → green → refactor when tests cover the touched area.
  • Use `EnterPlanMode` for test-driven planning before implementation.
  • If tests are absent or out of scope, state that explicitly.
Read more
Ships withswarm-tools

Multi-agent coordination that survives context death. Break big tasks into small ones. Spawn agents to work in parallel. Learn from what works. swarmtools.ai · Documentation

Get the whole plugin

Other skills on swarm-tools.