artifact-naming
Naming conventions for artifact kind and path, ensuring data can be aggregated and analyzed
wanman CLI command reference — inter-agent communication, task management, artifact storage
$ npx -y skills add chekusu/wanman --skill wanman-cli --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/wanman-cliContext preview
The summary Claude sees to decide when to auto-load this skill.
wanman CLI command reference — inter-agent communication, task management, artifact storage
name: wanman-cli description: wanman CLI command reference — inter-agent communication, task management, artifact storage
You are running inside the wanman Agent Matrix. Use the following CLI commands to collaborate with other agents.
On every startup, execute in order:
1. `cat ./CLAUDE.md` — Read your role guide 2. `wanman recv` — Check for pending messages 3. `wanman task list --assignee $WANMAN_AGENT_NAME` — View tasks assigned to you
| Command | Description | |---------|-------------| | `wanman send <agent> "<message>"` | Send a normal message | | `wanman send <agent> "<message>" --steer` | Urgent message (interrupts the target agent's current work) | | `wanman send human --type decision "<message>"` | Ask the human to make a decision | | `wanman send human --type blocker "<message>"` | Tell the human what is blocking progress | | `wanman recv` | Receive pending messages | | `wanman escalate "<message>"` | Escalate to the CEO agent |
**Note**: `--steer` forcefully interrupts the target agent's process and restarts it. Use only in urgent situations. Prefer explicit `--type decision|blocker` when sending to `human`, so the product can render the right card.
wanman task create "<title>" --assign <agent> [--priority 1-10] [--after id1,id2] [--desc "<description>"] wanman task list [--status <pending|in_progress|done|blocked>] [--assignee <agent>] wanman task get <id> wanman task update <id> --status <in_progress|blocked|done> [--result "<text>"] wanman task done <id> "<result>"
wanman artifact put --kind <kind> --path "<domain>/<category>/<item>" \ --source "<source>" --confidence <0-1> [--task <id>] [--file <path>] '<metadata json>' wanman artifact list [--agent <a>] [--kind <k>] wanman artifact get <id>
wanman context set <key> "<value>" wanman context get <key>
Used to share blueprints, summaries, and other information needed across agents.
wanman hypothesis create "<title>" [--rationale "<text>"] wanman hypothesis list [--status <proposed|active|validated|rejected>] wanman hypothesis update <id> --status <validated|rejected> [--outcome "<text>"] [--evidence <artifact-ids>]
wanman agents # List all agents and states wanman agents spawn <template> [name] # Spawn a clone (e.g. feedback-2) wanman agents destroy <name> # Destroy a dynamic clone
1. `wanman task update <id> --status in_progress` 2. Do the work, write outputs to the `./output/` directory 3. `wanman task done <id> "completion description"` 4. `wanman send ceo "Task complete: <title>, output: output/<filename>"`
| Variable | Description | |----------|-------------| | `WANMAN_AGENT_NAME` | Your agent name (set automatically) | | `WANMAN_URL` | Supervisor address (default `http://localhost:3120`) |
Agent Matrix framework — run a supervised network of Claude Code or Codex agents that collaborate on your machine. wanman is an open-source local-mode agent matrix framework.
Repo: chekusu/wanman
Naming conventions for artifact kind and path, ensuring data can be aggregated and analyzed
Quality standards for structured deliverables, ensuring data is verifiable and traceable
CEO checks consistency of agent outputs during post-completion review
Methodology for market research and data collection, ensuring data quality and source traceability
File and output conventions for agent workspaces