agent-audit
Audit agents spawned in the current/last run against the agent-selection taxonomy
Deploy an MCP server generated by forge-from-openapi --target=mcp-server; detect transport, deploy, smoke-test, print .claude.json snippet
> /plugin marketplace add heymegabyte/claude-skillsHow it fires
How this command gets triggered: by you, by Claude, or both.
/deploy-forged-mcpContext preview
What this command does when you run it.
Deploy an MCP server generated by forge-from-openapi --target=mcp-server; detect transport, deploy, smoke-test, print .claude.json snippet
description: Deploy an MCP server generated by forge-from-openapi --target=mcp-server; detect transport, deploy, smoke-test, print .claude.json snippet argument-hint: <mcp-server-dir> allowed-tools: Bash, Read, Write, Edit
Deploy a forged MCP server. Detects transport from `wrangler.toml` presence, deploys via Wrangler (http) or builds locally (stdio), smoke-tests with MCP Inspector, and outputs a ready-to-paste `.claude.json` registration block.
**When to use** — immediately after `/forge-from-openapi` produces an `mcp-server-dir`; when redeploying after schema changes.
**Inputs** — `$ARGUMENTS`: relative or absolute path to the MCP server directory.
---
MCP_DIR="${ARGUMENTS%/}" # strip trailing slash
ls "$MCP_DIR" 2>/dev/null || echo "NOTFOUND"---
ls "$MCP_DIR/wrangler.toml" "$MCP_DIR/wrangler.jsonc" 2>/dev/null | head -1
---
Only when `TRANSPORT=http`:
cd "$MCP_DIR" && npm install 2>&1
cd "$MCP_DIR" && npx wrangler deploy 2>&1
---
Only when `TRANSPORT=stdio`:
cd "$MCP_DIR" && npm install 2>&1
If `scripts.build` exists:
cd "$MCP_DIR" && npm run build 2>&1
ls "$MCP_DIR/dist/index.js" 2>/dev/null \ || ls "$MCP_DIR/build/index.js" 2>/dev/null \ || ls "$MCP_DIR/index.js" 2>/dev/null \ || echo "ENTRY_MISSING"
---
**http transport:**
npx --yes @modelcontextprotocol/inspector \ --transport http \ --url "$DEPLOYED_URL/mcp" \ --json 2>&1 | head -60
**stdio transport:**
npx --yes @modelcontextprotocol/inspector \ --transport stdio \ --command "node $ENTRY_PATH" \ --json 2>&1 | head -60
---
Read `~/.claude.json` for indentation shape; use default if absent.
**http:**
{
"mcpServers": {
"<server-name>": {
"type": "http",
"url": "<DEPLOYED_URL>/mcp"
}
}
}**stdio:**
{
"mcpServers": {
"<server-name>": {
"type": "stdio",
"command": "node",
"args": ["<ABSOLUTE_ENTRY_PATH>"]
}
}
}`<server-name>` = `name` from `package.json` (strip `@scope/` prefix). Print labelled:
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Paste into ~/.claude.json › mcpServers: ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ <json block> ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ After adding: run /audit-mcp-fleet --id <server-name> to confirm registration.
---
Deploy Summary — <server-name> Transport: http | stdio Deployed URL: https://... (http only) Entry point: /path/to/dist/index.js (stdio only) Build: ✓ | ✗ Smoke test: ✓ 12 tools | ✗ failed Next step: paste .claude.json snippet above, then restart Claude Code
On any failure: emit `✗ Deployment incomplete — see errors above before registering.`
---
**Verification** — After pasting and restarting, run `/audit-mcp-fleet --id <server-name>`.
**See**
14-category autonomous product-building OS for 32+ AI coding tools. One-line prompts → deployed products.
Repo: heymegabyte/claude-skills
Audit agents spawned in the current/last run against the agent-selection taxonomy
Run the Agent Diversity Review gate and emit the result table
Meta-analyze the effectiveness of a /loop arc — per-iteration metrics, LOC delta trend, saturation detection, and a keep/lengthen/delete recommendation.
Audit the rules/ directory for missing foundational principles; output gap list with priority and justification
Validate ~/.claude/settings.json hooks block — event names, file existence, executability, matcher syntax; --fix repairs common issues
Catch Resend-class bug (isError: false on HTTP 4xx/5xx) across all MCP server tool handlers