Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
sankalpgunturi avatar

Collab Claw Spike B

  • Updated April 27, 2026
  • sankalpgunturi/collab-claw-spike-b

collab-claw-spike-b is a Claude Code skill in the AI & Agent Building category. Spike B — verify the production-path host monitor (on-skill-invoke:host) wakes idle Claude on '[Name]: …' lines and that SKILL-body host instructions persist across the session.

Key points

  • collab-claw-spike-b
  • AI & Agent Building
  • AI-coding skill

Collab Claw Spike B by the numbers

  • Data as of Jul 7, 2026 (Skillselion catalog sync)
/plugin marketplace add sankalpgunturi/collab-claw-spike-b
/plugin install collab-claw-spike-b@collab-claw-spike-b

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Last updatedApril 27, 2026
Repositorysankalpgunturi/collab-claw-spike-b

What it does

Spike B — verify the production-path host monitor (on-skill-invoke:host) wakes idle Claude on '[Name]: …' lines and that SKILL-body host instructions persist across the session.

README.md

collab-claw — Spike B

A throwaway plugin used to verify a single thing: a plugin distributed via the Claude Code public marketplace can wake an idle Claude session via a plugin monitor scoped to on-skill-invoke:host, with [Name]: <text> lines being interpreted as new user requests from named teammates — using exactly the production architecture (skill body holds host instructions, no separate agent file, no settings.json agent default).

This is the real gate. If Spike B passes, the v1 architecture in PLAN.md is buildable.

Layout (matches production marketplace shape exactly):

collab-claw-spike-b/
├── .claude-plugin/
│   └── marketplace.json          ← repo-root marketplace catalog
└── plugin/
    ├── .claude-plugin/
    │   └── plugin.json           ← plugin manifest
    ├── skills/
    │   └── host/
    │       └── SKILL.md          ← name: host, disable-model-invocation: true
    │                               body holds host instructions + bash command
    ├── monitors/
    │   └── monitors.json         ← top-level array; when: on-skill-invoke:host
    └── bin/
        ├── host-noop             ← bash command the SKILL runs (prints banner)
        └── spike-b-emit          ← monitor body (time-emit + trigger watch)

What spike-b-emit does

Phase 1 — time-emit (deterministic):

  • t=30s → [Sankalp]: please write a hello-world Python script
  • t=60s → [Abhinav]: now translate that to Rust

Phase 2 — trigger-file watch (manual): polls ${CLAUDE_PLUGIN_DATA}/triggers/ every 2s.

  • charlie.fire → emits [Charlie]: please add a docstring to whatever you wrote last
  • burst.fire → emits [Dana]: … then 2s later [Eve]: …
  • quit → monitor exits cleanly

monitor.log next to the trigger dir shows pid, env vars, every emit, and lifecycle events.

Run the spike

# Terminal 1 — install the plugin in claude
claude
> /plugin marketplace add sankalpgunturi/collab-claw-spike-b
> /plugin install collab-claw-spike-b@collab-claw-spike-b
> /reload-plugins
> /collab-claw-spike-b:host        # starts the monitor

# Wait ~30s. [Sankalp]: ... should appear and Claude should respond as if Sankalp asked it.
# Wait another ~30s. [Abhinav]: ... should appear and Claude should answer Abhinav.

# Terminal 2 — manual triggers (after phase 1 completes)
touch ~/.claude/plugins/data/collab-claw-spike-b-collab-claw-spike-b/triggers/charlie.fire
# (path is also printed by /collab-claw-spike-b:host)

Test cases and pass/fail criteria are in SPIKE_B_RESULTS.md at the root of the parent repo.

Cleanup

claude
> /plugin uninstall collab-claw-spike-b@collab-claw-spike-b
> /plugin marketplace remove collab-claw-spike-b

# Then locally
touch ~/.claude/plugins/data/collab-claw-spike-b-collab-claw-spike-b/triggers/quit

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.