An MCP server that provides on-demand skill loading for AI coding assistants.
BY modbenderUPDATED Oct 28, 2026
INTENDED FOR◈OpenClawInstall in a listed agent.
instructions.md
---
name: banking-ae4db25bc6ec
slug: banking
version: 1.0.0
homepage: https://clawic.com/skills/banking
description: Manage retail and business banking workflows with payment operations, account controls, reconciliation, fraud response, and compliant communication.
changelog: Initial release with structured banking workflows for intake, operations, incident response, and customer-safe messaging.
metadata: {"clawdbot":{"emoji":"B","requires":{"bins":[],"config":["~/banking/"]},"os":["darwin","linux","win32"],"configPaths":["~/banking/"]}}
---
## Setup
On first use, read `setup.md` for activation boundaries and context capture priorities.
## When to Use
Use this skill for banking operations support: account onboarding workflows, payment operations, reconciliation triage, fraud incidents, and customer communication that must stay clear and compliant.
## Architecture
Memory lives in `~/banking/`. See `memory-template.md` for structure and status fields.
```text
~/banking/
|-- memory.md # Status, activation scope, operating context
|-- incidents.md # Open fraud and operations incidents
|-- payment-controls.md # Verified controls by rail and account type
`-- communication-notes.md # Approved customer messaging patterns
```
## Quick Reference
Use the smallest relevant file for the task to keep decisions precise under time pressure.
| Topic | File |
|-------|------|
| Setup process | `setup.md` |
| Memory template | `memory-template.md` |
| Request intake and classification | `intake-checklist.md` |
| Payment rails and controls | `payment-ops.md` |
| Fraud and outage handling | `incident-response.md` |
| Customer-safe wording | `customer-messaging.md` |
| Regulatory and legal boundaries | `compliance-scope.md` |
## Core Rules
### 1. Classify the Request Before Giving Steps
- Label each request first: onboarding, payment execution, reconciliation, fraud, dispute, or compliance question.
- If the category is unclear, ask one short clarification before proposing actions.
### 2. Confirm Jurisdiction and Account Context
- Capture country or region, customer type (consumer or business), and account type before compliance-sensitive guidance.
- Never give jurisdiction-specific legal conclusions without explicit location context.
### 3. Use Control-First Payment Guidance
- For every transfer path, verify account ownership, amount, cutoff timing, approval threshold, and rollback options.
- If any required control is unknown, pause execution advice and request the missing control.
### 4. Treat Incidents as Containment Then Recovery
- For suspected fraud or unauthorized activity, prioritize containment actions before root-cause analysis.
- Keep incident actions timestamped and reversible where possible.
### 5. Keep Communication Clear, Neutral, and Accurate
- Use plain language that states current status, next step, owner, and ETA window.
- Avoid guarantees, blame language, or speculative claims about pending investigations.
### 6. Keep Memory Actionable and Verifiable
- Record only durable context: operating boundaries, approved controls, known constraints, and recurring failure patterns.
- Do not store full account numbers, authentication data, or sensitive personal identifiers in memory notes.
### 7. Escalate High-Risk or Restricted Requests
- Escalate when requests involve sanctions, KYC circumvention, legal interpretation, or irreversible fund movement without controls.
- Refuse instructions that circumvent required approvals, customer consent, or regulatory safeguards.
## Common Traps
- Starting with product explanations instead of request classification -> slower resolution and wrong workflow.
- Giving transfer steps before confirming controls -> elevated operational and fraud risk.
- Mixing legal interpretation with operations guidance -> compliance exposure and user confusion.
- Responding to incidents with generic advice only -> delayed containment and larger losses.
- Using absolute language such as "guaranteed" or "always" -> credibility and regulatory risk.
- Logging sensitive data in memory notes -> avoidable privacy and security exposure.
## Data Storage
- Local notes only in `~/banking/` (memory file, incident notes, and control references).
- Keep stored content minimal and operational: controls, status, and decisions.
- Do not store full account numbers, authentication data, or unnecessary personal identifiers.
## Security & Privacy
Data that leaves your machine:
- None by default. This skill is instruction and workflow guidance only.
Data that stays local:
- Operational context and notes in `~/banking/`.
This skill does NOT:
- Access bank portals or execute fund transfers automatically.
- Request undeclared network calls.
- Store authentication data or full account numbers in memory files.
- Modify files outside `~/banking/` for storage.
- NEVER modifies its own skill definition file.
## Related Skills
Install with `clawhub install <slug>` if user confirms:
- `payments` - Payment workflows and transaction operations patterns.
- `accounting` - Ledger, reconciliation, and financial reporting support.
- `invoice` - Invoice lifecycle workflows and settlement tracking.
- `money` - Personal money management and budgeting fundamentals.
- `invest` - Investment analysis workflows for portfolio decisions.
## Feedback
- If useful: `clawhub star banking`
- Stay updated: `clawhub sync`
---
name: banking-ae4db25bc6ec
slug: banking
version: 1.0.0
homepage: https://clawic.com/skills/banking
description: Manage retail and business banking workflows with payment operations, account controls, reconciliation, fraud response, and compliant communication.
changelog: Initial release with structured banking workflows for intake, operations, incident response, and customer-safe messaging.
metadata: {"clawdbot":{"emoji":"B","requires":{"bins":[],"config":["~/banking/"]},"os":["darwin","linux","win32"],"configPaths":["~/banking/"]}}
---
## Setup
On first use, read `setup.md` for activation boundaries and context capture priorities.
## When to Use
Use this skill for banking operations support: account onboarding workflows, payment operations, reconciliation triage, fraud incidents, and customer communication that must stay clear and compliant.
## Architecture
Memory lives in `~/banking/`. See `memory-template.md` for structure and status fields.
```text
~/banking/
|-- memory.md # Status, activation scope, operating context
|-- incidents.md # Open fraud and operations incidents
|-- payment-controls.md # Verified controls by rail and account type
`-- communication-notes.md # Approved customer messaging patterns
```
## Quick Reference
Use the smallest relevant file for the task to keep decisions precise under time pressure.
| Topic | File |
|-------|------|
| Setup process | `setup.md` |
| Memory template | `memory-template.md` |
| Request intake and classification | `intake-checklist.md` |
| Payment rails and controls | `payment-ops.md` |
| Fraud and outage handling | `incident-response.md` |
| Customer-safe wording | `customer-messaging.md` |
| Regulatory and legal boundaries | `compliance-scope.md` |
## Core Rules
### 1. Classify the Request Before Giving Steps
- Label each request first: onboarding, payment execution, reconciliation, fraud, dispute, or compliance question.
- If the category is unclear, ask one short clarification before proposing actions.
### 2. Confirm Jurisdiction and Account Context
- Capture country or region, customer type (consumer or business), and account type before compliance-sensitive guidance.
- Never give jurisdiction-specific legal conclusions without explicit location context.
### 3. Use Control-First Payment Guidance
- For every transfer path, verify account ownership, amount, cutoff timing, approval threshold, and rollback options.
- If any required control is unknown, pause execution advice and request the missing control.
### 4. Treat Incidents as Containment Then Recovery
- For suspected fraud or unauthorized activity, prioritize containment actions before root-cause analysis.
- Keep incident actions timestamped and reversible where possible.
### 5. Keep Communication Clear, Neutral, and Accurate
- Use plain language that states current status, next step, owner, and ETA window.
- Avoid guarantees, blame language, or speculative claims about pending investigations.
### 6. Keep Memory Actionable and Verifiable
- Record only durable context: operating boundaries, approved controls, known constraints, and recurring failure patterns.
- Do not store full account numbers, authentication data, or sensitive personal identifiers in memory notes.
### 7. Escalate High-Risk or Restricted Requests
- Escalate when requests involve sanctions, KYC circumvention, legal interpretation, or irreversible fund movement without controls.
- Refuse instructions that circumvent required approvals, customer consent, or regulatory safeguards.
## Common Traps
- Starting with product explanations instead of request classification -> slower resolution and wrong workflow.
- Giving transfer steps before confirming controls -> elevated operational and fraud risk.
- Mixing legal interpretation with operations guidance -> compliance exposure and user confusion.
- Responding to incidents with generic advice only -> delayed containment and larger losses.
- Using absolute language such as "guaranteed" or "always" -> credibility and regulatory risk.
- Logging sensitive data in memory notes -> avoidable privacy and security exposure.
## Data Storage
- Local notes only in `~/banking/` (memory file, incident notes, and control references).
- Keep stored content minimal and operational: controls, status, and decisions.
- Do not store full account numbers, authentication data, or unnecessary personal identifiers.
## Security & Privacy
Data that leaves your machine:
- None by default. This skill is instruction and workflow guidance only.
Data that stays local:
- Operational context and notes in `~/banking/`.
This skill does NOT:
- Access bank portals or execute fund transfers automatically.
- Request undeclared network calls.
- Store authentication data or full account numbers in memory files.
- Modify files outside `~/banking/` for storage.
- NEVER modifies its own skill definition file.
## Related Skills
Install with `clawhub install <slug>` if user confirms:
- `payments` - Payment workflows and transaction operations patterns.
- `accounting` - Ledger, reconciliation, and financial reporting support.
- `invoice` - Invoice lifecycle workflows and settlement tracking.
- `money` - Personal money management and budgeting fundamentals.
- `invest` - Investment analysis workflows for portfolio decisions.
## Feedback
- If useful: `clawhub star banking`
- Stay updated: `clawhub sync`
COMPATIBILITY & REFERENCES
Platform guidance
Imported from the listed AgentSkill.sh category. Complete supplied package files and notices are retained. Required external tools and credentials are not installed. No bundled code was executed during import. Individual agent task behavior has not been tested.
Generate detailed control implementation guidance, technical steps, and implementation plans for OSCAL security controls. Use this skill to create implementation narratives, technical procedures, and deployment plans.
Soc2 Compliance Checker - Auto-activating skill for Security Advanced. Triggers on: soc2 compliance checker, soc2 compliance checker Part of the Security Advanced skill category.
The AI native file format. EXIF for AI — stamps every file with trust scores, source provenance, and compliance metadata. Embeds into 20+ formats (DOCX, PDF, images, code). EU AI Act, SOX, HIPAA auditing.
Assess worker-classification and compliance risk for temporary event staffing in the US and Canada — W-2 vs 1099, misclassification penalties, joint-employer liability, COI, and wage/hour rules. Includes live state-by-state lookups via MCP.