agentsclimarketplace

Smcp status

Skill Heretyc/subagent-mcp/skills/smcp-status

Report live subagent-mcp session state by calling the get_status MCP tool and rendering it in a uniform plain-ASCII layout. Use when the user says "smcp status", "/smcp:status", "routing status", "which providers active", or "agent count". Live data only; reads no state files.From its SKILL.md

Install
npx -y skills add Heretyc/subagent-mcp --skill smcp-status

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 3 stars3 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

3.4 KB, as published. Nobody here has run it

subagent-mcp Status

When triggered, call the get_status MCP tool and present its live session state. Use only what get_status returns. Do not read state files, guess, or cache: the numbers come from the running MCP server's in-memory state, so a fresh call each time is the single source of truth. If the get_status tool is not available, say the subagent-mcp MCP server is not connected in this session and point the user at the /smcp:doctor skill.

Data source

get_status returns a JSON object with these fields:

  • providers_loaded - array of provider names parsed from providers.jsonc at server start.
  • agent_count - number of agents launched since this server process started.
  • session_start_time - ISO timestamp of the MCP server process boot (may be null), used to compute uptime.
  • last_routing_decisions - array of { category, provider, timestamp, elapsed_ms }, most recent last, capped at the last 10.
  • swarm - agentic-swarm session snapshot: { active, current_stage, stage_name, pin_active, pin_expires_at }. When active is false all other swarm fields are null or false.

Presentation (E1-standardized)

Render uniform, one-item-per-line, plain-ASCII output. No smart quotes, no emoji, no tables. Use this exact shape:

subagent-mcp status
providers_loaded (N): name1, name2, ...
agent_count: K (since session start)
last routing decisions (most recent last):
  1. category=<category> provider=<provider> elapsed_ms=<n>
  2. category=<category> provider=<provider> elapsed_ms=<n>
  3. category=<category> provider=<provider> elapsed_ms=<n>

Rules for filling it in:

  • providers_loaded: print the count then the comma-separated names. If the array is empty, print providers_loaded (0): none loaded. If you can determine reachability (for example from a recent /smcp:doctor run in this conversation), add a short parenthetical note per provider such as (reachable) or (unreachable); otherwise omit the note rather than guessing - get_status alone does not probe reachability.
  • agent_count: print the integer verbatim. When session_start_time is present you may add uptime, for example agent_count: 4 (since session start, up 12m); if it is null, just print the count.
  • last routing decisions: show the last 3 entries from last_routing_decisions (the 3 most recent). Number them 1-3 with the oldest of the three first and the newest last. If fewer than 3 exist, show what is there. If the array is empty, print last routing decisions: none yet.

Add a swarm line after the routing decisions block:

  • When swarm.active is true: swarm: stage <current_stage> of 7 (<stage_name>), pin <active|off>
  • When swarm.active is false: swarm: idle

pin reflects swarm.pin_active: print active when true, off when false. Do not print pin_expires_at unless it adds useful context (for example, if the user asks how long the pin remains active).

Keep every line plain and aligned by field name so successive status reports read the same way.

Keep looking

Skills are one crate of 609,856. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.