agentleFS
tjg37tjg37/ironclaw/.claude/skills/add-agent/SKILL.md

add-agent

Skill1 starsChanged 7 months agoMITView on GitHub

Create a new IronClaw agent with its own persona, config, and model

  • Reads credentials

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

11 rules. 0 are common, 8 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 7 edited versions we read keep these.

  • 8

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Create an additional agent alongside the default one.

    line 8
  2. Very specific:

    IMPORTANT: Use the AskUserQuestion tool for all selections.

    line 10
  3. Distinctive:

    Ask the user for a name using AskUserQuestion with a few suggestions plus "Other":

    line 14
  4. Distinctive:

    Use AskUserQuestion (single select):

    line 23
  5. Distinctive:

    If they select "Other", show the remaining options:

    line 29
  6. Distinctive:

    Use AskUserQuestion (multi-select) with question "Which capabilities should be ENABLED?":

    line 47
  7. Very specific:

    All should be selected by default.

    line 53
  8. Distinctive:

    Use AskUserQuestion (multi-select) with question "Which integrations do you want?":

    line 57
  9. Distinctive:

    Run from the shared package directory:

    line 68
  10. Distinctive:

    Conversationally: Ask the current agent to "list all agents" to see it

    line 97
  11. Very specific:

    Configure: Run /configure to modify the agent's settings

    line 98

“Common” is measured: kept by at least 6 in 10 of the 7 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

/add-agent — Create a New Agent

Create an additional agent alongside the default one. Each agent has its own persona, boundaries, memory, and optionally a different model.

IMPORTANT: Use the AskUserQuestion tool for all selections.

Step 1: Name

Ask the user for a name using AskUserQuestion with a few suggestions plus "Other":

  • research-assistant — "For research and analysis tasks"
  • code-reviewer — "For code review and debugging"
  • content-writer — "For writing and editing"

The user can type a custom name via "Other".

Step 2: Persona

Use AskUserQuestion (single select):

  • General assistant — "Good at everything"
  • Developer assistant — "Code, debugging, architecture"
  • Research assistant — "Thorough research, citations, analysis"
  • Personal organizer — "Tasks, scheduling, reminders"

If they select "Other", show the remaining options:

  • Writing assistant — "Content creation, editing"
  • Data analyst — "Data exploration, SQL, visualizations"
  • DevOps / SRE — "Infrastructure, deployment, monitoring"
  • Product manager — "Requirements, specs, user stories"

They can also type a custom persona via "Other" on either question.

Step 3: Model

Use AskUserQuestion (single select):

  • Default (same as main agent) — "Uses the model from ANTHROPIC_MODEL env var"
  • Claude Haiku — "Fast and cheap, good for simple tasks (claude-haiku-4-5-20251001)"
  • Claude Sonnet — "Balanced performance (claude-sonnet-4-20250514)"
  • Claude Opus — "Most capable (claude-opus-4-20250514)"

Step 4: Boundaries

Use AskUserQuestion (multi-select) with question "Which capabilities should be ENABLED?":

  • Run bash commands — "Execute shell commands"
  • Write/edit files — "Create and modify files"
  • Web search — "Search the internet for information"
  • Read system files — "Access files outside the project directory"

All should be selected by default. Items the user deselects become false in boundaries.

Step 5: MCP Connections

Use AskUserQuestion (multi-select) with question "Which integrations do you want?":

  • Memory (built-in) — "Persistent long-term memory (recommended)"
  • GitHub — "Read/write GitHub issues, PRs, code. Requires GITHUB_TOKEN in .env."
  • Sentry — "Read Sentry issues/events. Requires SENTRY_AUTH_TOKEN in .env."
  • Slack — "Coming soon"
  • Google Calendar — "Coming soon"

If the user selects GitHub or Sentry, check .env for the matching token (GITHUB_TOKEN / SENTRY_AUTH_TOKEN). If missing, tell the user the token needs to be added before the agent will actually use that MCP — IronClaw will log a warning and skip the server at runtime. Do NOT block agent creation; the connection stays in config and starts working the moment the token is set.

Step 6: Create

Run from the shared package directory:

cd packages/shared && node --env-file=../../.env --import tsx/esm -e "
import { agentRepo } from './src/index.js';
const tenant = await agentRepo.findOrCreateDefaultTenant();
await agentRepo.createAgent(tenant.id, 'AGENT_NAME', {
  persona: 'CHOSEN_PERSONA',
  customPersona: 'CUSTOM_OR_UNDEFINED',
  boundaries: {
    allowBash: BOOL,
    allowFileWrites: BOOL,
    allowWebSearch: BOOL,
    allowSystemFiles: BOOL
  },
  mcpConnections: MCP_CONNECTIONS_ARRAY,
  model: 'MODEL_OR_UNDEFINED'
});
console.log('Agent created!');
process.exit(0);
"

Replace placeholder values with the user's actual choices. Omit customPersona and model if not applicable.

Step 7: Confirm

Show the created agent's config. Explain how to use it:

  • CLI: pnpm start -- --agent AGENT_NAME
  • Conversationally: Ask the current agent to "list all agents" to see it
  • Configure: Run /configure to modify the agent's settings

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.