CRM

Guide users building a personal CRM from simple files to structured database.

Security Scan

Instructions are coherent with a personal-CRM-from-files goal: no unexpected credentials, no external installs, and the runtime steps (create ~/crm, offer scripts/migrations) match the described purpose.

Purpose & Capability

The skill's name and description (personal CRM from files → structured DB) align with its contents: it focuses on file-based schemas, when to add interaction/company/deals files, and when to migrate to SQLite. There are no unrelated required env vars, binaries, or installs.

Instruction Scope

SKILL.md instructs the agent to create a ~/crm/ folder, propose helper scripts, and offer to write migration scripts. Those actions are coherent with the purpose but mean the agent may create and modify files and generate code to run locally — the user should review any generated scripts before executing them. The instructions do not request access to unrelated system paths or external endpoints.

Install Mechanism

No install spec and no code files — instruction-only. This is low risk: nothing will be downloaded or written by an installer. Any file writes would come from agent-suggested commands or generated scripts at runtime.

Credentials

The skill requires no environment variables, credentials, or config paths. That is proportional for a local, single-user CRM workflow described in the doc.

Persistence & Privilege

always is false and the skill does not request persistent elevated privileges or modification of other skills or system-wide settings. Its recommended actions are limited to the user's home directory (~/crm) and optional helper scripts.

Assessment

This skill appears coherent and low-risk, but keep these precautions: 1) Review any commands or scripts the agent generates (especially migration scripts) before running them — they may modify or overwrite your files. 2) Back up ~/crm/ (or follow the skill's own backup advice) before bulk edits or migrations. 3) If the skill suggests cloud sync, consider the privacy implications of storing contacts in third-party services. 4) If you’re uncomfortable with automatic file writes, ask the agent to provide scripts/commands for manual review and execution instead of running them directly.

First Interaction

• Ask what they're tracking: clients, leads, investors, job contacts, networking — context shapes schema
• Ask their technical comfort: spreadsheets, JSON, databases — determines starting format
• Create ~/crm/ folder as the single source of truth

Start With Files, Not Apps

• JSON or CSV for first version — validate data model before adding complexity
• Single file contacts.json initially — resist creating multiple files until needed
• Don't suggest web app until they've used files for at least a week
• Don't suggest database until files feel slow or limiting

Minimal Contact Schema

• id, name, email, company, phone, notes, tags, created, updated — nothing more initially
• tags array over rigid categories — flexible, no schema changes needed
• notes field is often the most valuable — encourage freeform context
• Generate UUID for id, not auto-increment — survives merges and imports

When To Add Interactions File

• User asks "when did I last talk to X" — signal they need history
• Separate file linked by contact_id — not nested in contact object
• type field (note/email/call/meeting) enables filtering later
• Always include date — timeline view is essential

When To Add Companies File

• Multiple contacts at same company — signal to separate
• Many-to-many: one person can work at multiple companies over time
• company_id in contacts, not company name duplication

When To Add Deals/Opportunities

• User mentions "pipeline", "stage", "close date", "deal value"
• Link to contact_id and optionally company_id
• Stages as simple string field initially — don't over-engineer state machine

SQLite Migration Triggers

• File operations feel slow (>100 contacts typically)
• User wants to query/filter in complex ways
• Multiple users need access (SQLite handles concurrent reads)
• Offer to write migration script — don't force manual re-entry

Progressive Timeline

• Week 1: contacts file only, prove they'll use it
• Week 2: add interactions when they want history
• Week 3: add tags, search helper script
• Month 2: companies file if needed
• Month 3: deals file if tracking opportunities
• Only then: consider web UI or more complex tooling

What NOT To Suggest Early

• Web application — massive scope increase, validate data model first
• Email sync/integration — suggest BCC/forward workflow, much simpler
• Calendar integration — manual logging is fine initially
• Authentication — single-user local CRM doesn't need it
• Mobile app — sync complexity not worth it early

Helper Scripts Worth Offering

• Quick add from command line — reduces friction
• Search across all files — grep/jq one-liner
• Backup to timestamped zip — essential before migrations
• Export to CSV — for users who want spreadsheet view

Data Integrity Habits

• Backup before any bulk edit or migration
• Check for duplicate emails before adding contact
• Validate email format on entry
• Keep created/updated timestamps — debugging lifesaver

Sync When Asked

• Cloud folder (Dropbox/iCloud/Drive) for multi-device — simplest
• Git repo for version history — good for technical users
• Don't suggest complex sync solutions until files prove insufficient