agent-audit
Audit agents spawned in the current/last run against the agent-selection taxonomy
Batch-forge Brian's entire stack as skills + MCP servers in one prompt; skips already-forged entries; applies automatic pruning when tool count >100
> /plugin marketplace add heymegabyte/claude-skillsHow it fires
How this command gets triggered: by you, by Claude, or both.
/forge-stack-packContext preview
What this command does when you run it.
Batch-forge Brian's entire stack as skills + MCP servers in one prompt; skips already-forged entries; applies automatic pruning when tool count >100
description: Batch-forge Brian's entire stack as skills + MCP servers in one prompt; skips already-forged entries; applies automatic pruning when tool count >100 argument-hint: [--target=skill|mcp-server|both] [--dry-run] [--only=<api-name>] allowed-tools: Bash, Read, Write, Edit, Glob
Batch-forge every API in Brian's stack that doesn't yet have a skill or MCP server. Runs `bin/forge-skill-from-openapi.mjs` for each eligible target, prunes oversized MCP servers, and emits a summary table.
**When to use** — fresh machine setup; after adding a new API; to fill gaps after `/audit-mcp-fleet`.
**Inputs** (all optional)
---
API | OpenAPI URL | skill-dir | mcp-dir --------------|------------------------------------------------------------------------------------------|-------------------------------------|----------------------------------- stripe | skip — already forged | skills/stripe | mcp-servers/stripe-mcp github | skip — already forged | skills/github | mcp-servers/github-mcp square | skip — already forged | skills/square | mcp-servers/square-mcp twilio | skip — already forged | skills/twilio | mcp-servers/twilio-mcp resend | https://raw.githubusercontent.com/resend/resend-api-docs/main/openapi.yaml | skills/resend | mcp-servers/resend-mcp bitwarden | https://api.bitwarden.com/openapi.json | skills/bitwarden | mcp-servers/bitwarden-mcp openai | https://github.com/openai/openai-openapi/raw/master/openapi.yaml | skills/openai | mcp-servers/openai-mcp anthropic | skip — no public OpenAPI | — | — posthog | https://posthog.com/openapi.json | skills/posthog | mcp-servers/posthog-mcp cloudflare | skip — mcp-servers/cloudflare-mcp exists (partial batch 3) | skills/cloudflare | mcp-servers/cloudflare-mcp
**Curated prune allowlists** (auto-applied when tool count > 100 after forge):
---
TARGET=$(echo "$ARGUMENTS" | grep -oP '(?<=--target=)\w+' || echo "both") DRY_RUN=$(echo "$ARGUMENTS" | grep -q '\-\-dry-run' && echo "true" || echo "false") ONLY=$(echo "$ARGUMENTS" | grep -oP '(?<=--only=)\S+' || echo "")
For each non-`skip` API in the manifest:
1. Check skill dir: `ls skills/<api-name> 2>/dev/null` 2. Check MCP dir: `ls mcp-servers/<api-name>-mcp 2>/dev/null` 3. Determine missing targets per `--target` flag. 4. If `--only=<name>` set, filter to that API; unknown name → error + stop.
Emit work plan:
Forge Stack Pack — work plan ══════════════════════════════════════════════════════════ API skill exists mcp exists action resend ✗ ✗ forge skill + mcp-server bitwarden ✗ ✗ forge skill + mcp-server openai ✗ ✗ forge skill + mcp-server posthog ✗ ✗ forge skill + mcp-server cloudflare ✓ ✓ skip (already forged) stripe ✓ ✓ skip (already forged) github ✓ ✓ skip (already forged) square ✓ ✓ skip (already forged) twilio ✓ ✓ skip (already forged) anthropic — — skip (no public OpenAPI) ══════════════════════════════════════════════════════════ Total to forge: 4 APIs · 4 skills · 4 mcp-servers
If `--dry-run`: print table and stop.
---
Process APIs **sequentially** (not parallel) to avoid wrangler + npm contention.
API_NAME="<api-name>"
SPEC_URL="<url-from-manifest>"
SPEC_FILE="/tmp/forge-stack-${API_NAME}-spec.json"
curl -fsSL "$SPEC_URL" -o "$SPEC_FILE" 2>&1node -e "const y=require('js-yaml');const fs=require('fs');fs.writeFileSync('$SPEC_FILE'.replace('.json','-orig.yaml'),fs.readFileSync('$SPEC_FILE'));fs.writeFileSync('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