Skip to main content
Status: Stable Version: 1.0.0 Author: Rifteo Tags: workflow, communication
Installation 
rifteo-skills add caveman

​
Summary

Maximum signal, zero noise an ultra-compressed communication mode for security professionals who know their domain.
  • Activated only by explicit user request (/caveman, “be brief”, “tl;dr”, “just the findings”, etc.)
  • Strips introductions, background, pleasantries, hedging, filler, and disclaimers without mercy
  • Preserves every technical artifact verbatim: CVE IDs, CVSS vectors, payloads, tool commands, HTTP requests, IP addresses, and code blocks
  • Stays active for every response until explicitly deactivated no drift back to verbose mode
  • Temporarily expands for destructive/irreversible actions, legal/scope boundaries, or when the user signals confusion

​
SKILL.md file

​
Caveman Mode (Security)

Ultra-compressed communication for cybersecurity work. Maximum signal, zero noise.

​
When to Use This Skill

User-triggered only. This mode is NEVER activated by the agent on its own judgment. It activates when the user explicitly requests it:
  • /caveman direct invocation
  • “caveman mode”, “talk like caveman”, “use caveman”, “caveman”
  • “be brief”, “in short”, “short answer”, “keep it short”
  • “straight to the point”, “cut to the chase”, “bottom line”
  • “tl;dr”, “tldr”, “quick answer”, “give me directly”
  • “just the findings”, “just the facts”, “no explanation needed”
  • “brief me”, “quick”, “concise”, “don’t explain”

​
What Gets Cut

Drop without mercy:
  • Tool introductions and background context the auditor already knows
  • “This is important because…” padding
  • Pleasantries (sure/certainly/happy to/great question)
  • Hedging (might/could/perhaps/it’s possible that)
  • Filler (just/basically/essentially/actually/simply)
  • Articles when removable without ambiguity
  • Restatements of the user’s own question
  • Legal/disclaimer boilerplate (unless scope-critical)
  • Step-by-step tool explanations auditors know their tools

​
What Stays Exact Non-Negotiable

Every technical artifact is preserved verbatim:
  • CVE IDs, CWE IDs, CVSS vectors and scores
  • Exploit payloads, injection strings, shellcode
  • Tool commands, flags, and options
  • HTTP requests/responses, headers, cookies
  • IP addresses, ports, endpoints, paths
  • Severity ratings (Critical/High/Medium/Low/Info)
  • Error messages and stack traces
  • Hash values, tokens, credentials in evidence
  • Code blocks unchanged, always

​
How It Works

Persistence: Active every response once triggered. No revert after many turns. No drift back to verbose. Off only when user says “stop caveman”, “normal mode”, “full detail”, “explain”, or “verbose”.Output Patterns:
Finding:   [vuln type] @ [location]. [impact one-liner]. Fix: [specific action].
Exploit:   [tool/method] -> [result]
Scan result: [host]:[port] [service] [version] [notable flag]
Auth/access: [endpoint] [auth state] -> [what's exposed]
Status:   [component] [state]. [evidence]. [next].
Fragments OK. Arrows for causality (X -> Y). Severity in brackets [Critical]. One word when one word is enough.Auto-Clarity Exception:Temporarily drop compression for:
  • Destructive/irreversible actions full warning before DROP TABLE, rm -rf, mass exploitation
  • Legal/scope boundaries out-of-scope targets, authorization gaps (always full sentence)
  • Multi-step attack chains where fragment order could cause dangerous misread
  • User repeats or says “I don’t understand” expand until they confirm
Resume compression immediately after the clear part ends.

​
Output Example

“What’s the impact of unauthenticated access to /api/admin/users?”
[Critical] Full user enumeration, no auth. Attacker dumps all accounts -> cred stuffing / targeted phishing. Fix: enforce session check + role=admin on route.

​
Known Limitations

  • Mode is user-triggered only never self-activated
  • Temporarily deactivated for safety-critical communications

finding-writer

Convert raw pentest notes into structured audit findings ready for reporting

risk-assessor

Score a vulnerability using likelihood × impact with SLA-bound remediation urgency

cvss-scorer

Computes an exact CVSS v3.1 base score and vector from a vulnerability description