rossum-sa

🧰 Rossum toolkit for Claude Code

Turn Claude into a Rossum implementation partner — audit hooks, analyze schemas, query Data Storage, extract documents, and generate SOWs, all from your terminal.

7 skills · 9 reference packs · 53 MCP tools — Claude Code plugin marketplace for Rossum.ai.

🚀 Quick start

You need Claude Code CLI and a Rossum API token.

/plugin marketplace add rossumai/claude-marketplace
/plugin install rossum-sa@rossumai-claude-marketplace

Then connect and go:

Connect to Rossum (token: <TOKEN>, base URL: https://elis.rossum.ai).
Map out the entire org — workspaces, queues, hooks, schemas — and draw
an ASCII architecture diagram. Add emoji health indicators next to each
component (🟢 healthy, 🟡 warning, 🔴 broken).

Note: Auto-updates are off by default for third-party marketplaces. Enable them in /plugin → Marketplaces tab.

Optional: For a better experience when developing and implementing projects, install the superpowers plugin — it adds structured planning, TDD workflows, and code review skills that pair well with rossum-sa:

/plugin install superpowers@claude-plugins-official

Running from a local checkout (feature branch)

To test a feature branch before it's published to the marketplace, clone the repo and point Claude Code at the plugin directories with --plugin-dir.

Example using the teamd-hackathon branch:

git clone -b teamd-hackathon https://github.com/rossumai/claude-marketplace.git rossum-claude-plugin

claude --plugin-dir rossum-claude-plugin/plugins/rossum-sa \
       --plugin-dir rossum-claude-plugin/plugins/nerossum

Each --plugin-dir loads one plugin; omit the ones you don't need.

⚡ Skills

rossum-sa

Skill	Description
/rossum-sa:write-sow	Generate a Statement of Work from project requirements
/rossum-sa:analyze [path]	Check an implementation for configuration errors
/rossum-sa:document [path]	Produce a queue-focused reference document
/rossum-sa:implement	Plan and execute an integration project end-to-end
/rossum-sa:refine-deployment [deploy-file] [impl-path]	Enhance prd2 deploy files with target IDs and attribute overrides
/rossum-sa:upgrade [path]	Upgrade deprecated extensions to formula fields and bump old Python runtimes on function hooks to python3.12
/rossum-sa:dead-code [path]	Find unused hooks, formulas, rules, labels, and engines with a deterministic detector

nerossum

Skill	Description
/nerossum:document-processing	Extract structured data from invoices, POs, and receipts with validation and anomaly detection

📚 Autoloaded references

When rossum-sa is enabled, Claude automatically gets domain knowledge for:

• Rossum platform — queues, schemas, hooks, annotations, workflows
• MongoDB — query syntax, aggregation pipelines
• Master Data Hub (MDH) — matching, scoring, collections
• Data Storage API — CRUD, indexing, search
• TxScript & Serverless Functions — formula fields, extension development
• SAP Integration — connector setup, mapping
• Export Pipeline (Request Processor) — multi-stage API integration engine, SFTP export, auth, response handling
• Coupa Integration Baseline (CIB) — schema, MDH matching, export pipeline, business rules
• prd2 CLI — deployment and management commands
• Structured Formats Import (SFI) — XML/JSON import setup, XPath/JMESPath selectors, e-invoicing (ZUGFeRD, X-Rechnung)

💡 What can you do with this?

🕵️ Who's been busy? — Pull a year of audit logs and surface suspicious activity patterns.

Connect to Rossum (token: <TOKEN>, base URL: https://elis.rossum.ai), pull all audit logs
for the last year, and print a histogram of user activity. Highlight suspicious patterns.

🔗 Find broken hooks — Audit your hook chains across all queues.

Connect to Rossum and list all hooks. Group them by queue and flag any that are inactive,
have no queues attached, or have a broken run_after chain.

📊 Spot missing indexes — Catch Data Storage performance problems before they bite.

Connect to Rossum and check all Data Storage collections. List their indexes and search
indexes, flag any missing __dynamic_index or duplicate/redundant indexes.

🎯 Tune fuzzy matching — Optimize MDH search scores with real data.

Connect to Rossum and find the $search query in the MDH matching extension. Verify and
calibrate the score thresholds against real data in the collection. Use at least 1000 samples.