danxbot
v0.7.22Danxbot orchestrator domain: dispatch architecture, runtime modes, deploy. Install on danxbot-aware instances.
By Dan Newman0 GitHub starsUpdated 2 days ago
Directory evidence
- Runtimes
- Claude Code
- Parsed components
- 27 skill or MCP entries
- Source updated
- Sep 21, 2026
- Manifest status
- Canonical path parsed
The directory validates manifest shape and source location. It does not execute the plugin or provide a security endorsement. Review the indexing methodology →
Install danxbot for Claude Code
claude plugin marketplace add IchenDEV/agent-plugin-mkt
claude plugin marketplace update agent-plugin-marketplace
claude plugin install danxbot@agent-plugin-marketplacePaste and run these commands in a terminal with Claude Code. They add and refresh the PluginsMP catalog, then install this plugin.
The installer fetches third-party code from the source repository shown on this page. This directory validates manifest structure and source location, but does not perform a security audit; review the manifest, components, and source before installing.
Get the source manually
git clone https://github.com/newms87/claude-pluginsClone the source repository, then follow its setup instructions to add the plugin to a compatible client. The plugin root is danxbot/.
Plugin files
├── .claude-plugin/plugin.json├── skills/autonomous-mode/SKILL.md├── skills/comment-style/SKILL.md├── skills/danx-chat/SKILL.md├── skills/danx-epic-link/SKILL.md├── skills/danx-flesh-out/SKILL.md├── skills/danx-ideate/SKILL.md├── skills/danx-next/SKILL.md├── skills/danx-start/SKILL.md├── skills/danx-triage-card/SKILL.md├── skills/danx-triage-orchestrator/SKILL.md├── skills/danxbot/SKILL.md├── skills/db-reset/SKILL.md├── skills/dispatch-deep/SKILL.md├── skills/docker-deep/SKILL.md├── skills/git-add-on-create/SKILL.md├── skills/halt-flag/SKILL.md├── skills/issue-blocker/SKILL.md├── skills/issue-card-workflow/SKILL.md├── skills/no-false-blockers/SKILL.md├── skills/no-unauthorized-worker-launch/SKILL.md├── skills/plan-workflow/SKILL.md├── skills/prepare-for-compaction/SKILL.md├── skills/prod-access/SKILL.md├── skills/settings-deep/SKILL.md├── skills/slack-agent/SKILL.md├── skills/template-app-build/SKILL.md└── skills/unblock/SKILL.md
Included Skills27
Zero-interactive-prompt contract for dispatched workers: decide unilaterally, escalate via an open problem, never AskUserQuestion or plan-mode pause.
Markdown style contract for description/comments[]/retro fields the dashboard Issues drawer renders; never-do list for escapes and ASCII tables.
Per-card chat agent: single Claude session-per-card, one turn per dispatch, resumes via claude --resume; optional card edit via mcp__danx-dashboard__issue_* tools on explicit ask.
Wire two-way parent_id ↔ children[] linkage between an Epic and its existing phase cards. Never creates cards.
Per-card flesh-out agent: expands half-baked card to full spec, populates ac[], splits into phases when scope warrants. Read-only on repo.
Launch the ideator agent to explore the repo, build knowledge, and generate feature cards.
Pull top ToDo card and run full autonomous card-processing workflow. Orchestrator: decide unilaterally or escalate by opening a problem, never wait on human.
Process all ToDo issue cards sequentially using the autonomous workflow.
Per-card triage agent: reads ONE card, decides per status (Review/Blocked/Waiting On), records a confidence score via MCP.
Operator-invoked triage orchestrator: drains Review list via parallel batches of up to 3 in-session subagents. Single-shot.
Networking model, runtime envs, deployment vs local, issue-tracker poller boundary, MCP server consumers of danxbot dispatches.
Sanctioned destructive-DB-reset path: bash <worktree>/.danxbot/safe-reset-db.sh. Forbids replicating it via psql/artisan/raw SQL.
Deep dispatch contracts: resume protocol, staged_files validation, Playwright proxy binary-safety, usage dedup, stall + stream-idle recovery.
Deep contracts for danxbot container layout, root .mcp.json inject, .env.<target> overlay merge, clean-room cwd isolation.
Git-add-on-create discipline for dispatched agents: stage a new file immediately after creating it so autosave protects it.
Decision rule for status:"critical_failure" vs "failed" and the danxbot_complete signaling contract on env-level blockers.
Gating checklist for stamping blocked:{at,reason} on a card, and for escalating to a human via an open problem: false-blocker audit, forbidden git ops, programmatic substitute, root-cause trace.
Issue card lifecycle: status derivation, mcp__danx-dashboard__issue_* tools, comment/retro/blocked/waiting_on contracts, container-atomic rule (Epic + Feature). LOAD BEFORE proposing card TYPE or how to slice work into cards (epic/feature/story) — recommending a card structure in chat is the gated action, not just the issue_create call; the slice/taxonomy gate that decides epic-vs-story lives ONLY in this body, so naming a type before loading = deciding it wrong.
Three commonly-mistaken false-blocker patterns + their programmatic substitutes; forbidden destructive working-tree git ops list.
Strict per-invocation user-auth gate for make launch-worker*, make deploy*, and any command that starts a danxbot worker or prod deploy.
THE planning workflow for any task with a human in the loop beyond a quick cleanup — multi-step plan or build; ANY question whose answer affects a plan; monitoring over time; resuming or handing off unfinished work; operator pastes a plan link. The danxbot Plan is the ONLY planning record: goals/rules/caveats = plan records, design = architecture sections, work + operator questions = cards attached to the plan. No plan files, no `~/.claude/plans/*.md`, no repo `.md` specs, no HTML pages, no chat summaries. Operator question = `Task` card + open problem with solutions, never chat, never AskUserQuestion. Keep 3 cards in flight on own sub-agents; never ask permission to dispatch. Load before connecting to, creating, or writing a plan.
Operationalizes plan-workflow's zero-context rule into a mechanical checklist for the moment BEFORE a compaction — capture the real tree, attribute interleaved uncommitted work per card, sweep stale In Progress claims, resolve a stopped agent's uncommitted work by committing it (never discarding), route facts to record/note/card, then RE-VERIFY immediately before declaring ready. Load when: the operator asks to prepare for/wrap up before compaction; context is visibly running low and work is unfinished; handing off a session; about to stop with sub-agents still dispatched or a dirty tree. Does not cover ongoing plan hygiene (see danxbot:plan-workflow) or git safety rules (see dev:git-discipline) — cross-references both rather than repeating them.
Three reach paths into production (HTTP API / SSH / make targets), canonical debug recipes, forbidden-actions list.
<repo>/.danxbot/settings.json schema, ownership matrix, writer-merge invariants, isFeatureEnabled hot path.
Slack-worker dispatch contract: danxbot_slack_reply required for questions, forbidden for tasks; danxbot_complete always; emoji reaction + card-view update for tasks.
MANDATORY on every template-app dispatch. Owns the multi-app build loop — read the template id(s) from the task prompt, then per id call load_template_app → edit the three editable SFC files → save_template_app, and signal danxbot_complete when every app is shipped.
Unblock-report contract: extract blocker, what is done vs what operator must do, exact commands, success/failure branch.
Plugin manifests1
{
"name": "danxbot",
"description": "Danxbot orchestrator domain: dispatch architecture, runtime modes, deploy. Install on danxbot-aware instances.",
"version": "0.7.22",
"author": {
"name": "Dan Newman",
"email": "[email protected]"
}
}For maintainers
If you maintain this plugin, link to this source-backed listing from your README so users can review its manifest and indexed components.
[danxbot on Agent Plugins Marketplace](https://pluginsmp.com/plugins/danxbot)