"Don't send your grocery list to your dog walker. Don't tell your boss you're interviewing elsewhere. Don't leak trade secrets to the world."
Your personal message bouncer that stops you from nuking your life with one wrong message.
We've all been there:
- "Oops I sent that?!" moments
- Drafted a rant about your boss, sent it to... your boss
- Accidentally shared client secrets with competitors
- World seeing messages meant for your partner
One message can cost you:
- 💰 Your job ($50K-500K/year)
- 💔 Your relationships (priceless)
- ⚖️ Legal liability ($$$$$)
- 🏆 Your reputation (irreparable)
Context Guard validates EVERY message before it leaves your hands. Like a friend grabbing your phone at 2am before you text your ex.
# Before sending to group chat:
./scripts/validate-message.sh "Your message" "channel-id"
✅ SAFE → Send away!
❌ VIOLATION → FIX IT FIRST# 1. Install the guard
./install.sh
# 2. Set up your contexts
./scripts/setup-context.sh \
"work" \
"telegram:-100123456" \
"projects,meetings" \
"gossip,complaints,side-hustle"
# 3. Validate before EVERY group message
./scripts/validate-message.sh \
"Q3 deliverables on track" \
"telegram:-100123456"
# Output: ✅ SAFE TO POST| Message | Destination | Guard Says |
|---|---|---|
| "My boss is an idiot" | #work-general | ❌ BLOCKED: "boss" + "idiot" in work context |
| "Client A is paying $50K" | #competitor-chat | ❌ BLOCKED: Client data in competitor context |
| "I'm pregnant" | Current boss DM | ❌ BLOCKED: TMI in professional context |
| "Our revenue is $847K" | Public Twitter | ❌ BLOCKED: Financial data in public context |
| "I hate Karen" | Group with Karen | ❌ BLOCKED: Gossip about person in their presence |
Your Message → Context Guard → [VALIDATION] → Channel
↓
Matches context rules? → YES → Send! ✅
Forbidden topics? → NO → BLOCKED! 🛑
skills/context-separation/
├── install.sh # Setup wizard
├── contexts.json # Your context rules
├── scripts/
│ ├── validate-message.sh # Message validator
│ ├── enforce-context.sh # Mandatory enforcer (exit 1 on violation)
│ ├── setup-context.sh # Create new contexts
│ └── review-violations.sh # Learn from mistakes
└── violations.log # Your disaster prevention log
Define your communication "worlds":
{
"work": {
"channels": ["telegram:-100123456"],
"allowed_topics": ["projects", "meetings"],
"forbidden_topics": ["gossip", "side-hustle"]
},
"side-hustle": {
"channels": ["telegram:-100789012"],
"allowed_topics": ["revenue", "strategy"],
"forbidden_topics": ["day-job", "current-employer"]
},
"family": {
"channels": ["telegram:-100345678"],
"allowed_topics": ["events", "love"],
"forbidden_topics": ["work-complaints", "exes"]
}
}- Multi-platform: Telegram, Slack, Discord, WhatsApp
- Violation logging: Track patterns to avoid
- Learning system: Gets smarter over time
- Zero exceptions: Must validate before sending
- Exit codes:
0= safe,1= violation (blocks CI/CD, scripts)
# In your scripts/workflows:
./scripts/enforce-context.sh "$MESSAGE" "$CHANNEL" || exit 1
# ^ If validation fails, EVERYTHING STOPSProfessional: "Context-aware message validation for multi-channel AI agents"
Fun: "Like having a friend grab your phone before you drunk-text your ex"
Dramatic: "One message can destroy your life. We stop that."
Corporate: "Enterprise-grade message governance with audit trails"
Gen Z: "POV: You almost sent 'I hate my boss' to your boss #saved"
Apache 2.0 — Save your career, fork it, share it.
Built with 💙 by humans who've learned the hard way.
"Trust is everything. One mistake kills it." — Chef's Law
#staycuriousANDkeepsmilin 🐧✨