Lab — Capstone
Execute capstone on `$AI_LAB` — checklist, lab path, evidence pack sealed; zero stranger access.
Visual · t32_capstone_lab
Capstone: checklist × $AI_LAB literacy evidence. Original Cyberlium.
Close the AI Security literacy loop — apps secured, guards evaluated, findings shipped, RoE signed — without jailbreak kits or stranger app tests.
Walk Topic 32 checklist on YOUR `$AI_LAB`: verify RoE, lab path current, LLM app and secure dev labs complete, governance and findings packs indexed, scan for jailbreak kits and stranger app refs (must be zero). Update index with capstone UTC completion. Integrity paragraph: YOUR apps discipline and authorized red team key insight from YOUR path. No jailbreak weaponization, no stranger app testing, no prod LLM without RoE, no fraud deepfake kits. Optional mentor review — three clarity questions. Rotate lab keys after if brief requires.
1. Execute checklist
Check each row; link artifact or N/A; confirm zero refused techniques in any file.
Scan sealed pack for bypass kits and stranger refs — remove before handoff.
Command guide
Try these commands — Execute checklist
═══ TOOLS & WEBSITES ═══ Browse / read these (authorized learning only — stay in YOUR lab / program scope)
OWASP LLM Top 10 — https://owasp.org/www-project-top-10-for-large-language-model-applications/ NIST AI RMF — https://www.nist.gov/itl/ai-risk-management-framework MITRE ATLAS — https://atlas.mitre.org/ OpenAI safety — https://openai.com/safety
═══ INSTALL ═══
Linux (Debian/Ubuntu):
Command — copy this
sudo apt install python3
macOS:
Command — copy this
brew install python3
Windows: Download https://python.org/downloads/
═══ LINUX / macOS ═══
Command — copy this
export AI_LAB=${AI_LAB:-$HOME/cyberlium-lab/t32-ai}
grep NEVER "$AI_LAB/notes/no-weaponized-ai.txt"
grep -E 'HARD BAN|NEVER' "$AI_LAB/threats/deepfake-detection-signals.md" "$AI_LAB/threats/ai-malware-named-literacy.txt" "$AI_LAB/findings/responsible-ai-disclosure.txt"
echo '=== Capstone: injection bot → OWASP → threats → eval → governance → findings ==='
python3 "$AI_LAB/bot/echo_bot.py" capstone-test
grep -E 'LLM0' "$AI_LAB/owasp/llm-top10-index.txt" | head -5
grep '\[ \]' "$AI_LAB/notes/capstone-checklist.txt" | head -5
python3 - <<'PY'
import os
lab = os.environ.get('AI_LAB', os.path.expanduser('~/cyberlium-lab/t32-ai'))
path = os.path.join(lab, 'notes', 'capstone-summary.txt')
open(path, 'w').write('
'.join([
'AI Security Capstone',
'scope: YOUR $HOME/cyberlium-lab/t32-ai/ + YOUR toy echo-bot',
'injection: direct/indirect named + YOUR bot tests only',
'poisoning: named risks + eval/supply chain checklists',
'threats: phishing/deepfake/malware detection signals only',
'owasp: LLM Top 10 mapped findings + app security checklists',
'secure dev: SDLC + eval + secrets hygiene',
'governance: NIST AI RMF aligned notes + authorized red team scope',
'never: malware gen / deepfake fraud / jailbreak stranger SaaS / stranger API scrape',
]))
print(f'Wrote {path}')
PYCommand — copy this
grep -E 'scope|never|owasp' "$AI_LAB/notes/capstone-summary.txt"
═══ WINDOWS ═══
Capstone on YOUR lab Windows VM only
Command — copy this
Get-Content $HOME/cyberlium-lab/t32-ai/notes/no-weaponized-ai.txt | Select-String NEVER Get-Content $HOME/cyberlium-lab/t32-ai/notes/capstone-summary.txt | Select-String never python $HOME/cyberlium-lab/t32-ai/bot/echo_bot.py capstone-test
Primary tools to practice this lesson: grep, python3. Reference sites: OWASP LLM Top 10 (https://owasp.org/www-project-top-10-for-large-language-model-applications/); NIST AI RMF (https://www.nist.gov/itl/ai-risk-management-framework); MITRE ATLAS (https://atlas.mitre.org/); OpenAI safety (https://openai.com/safety). Run every command in the box — install first, then the usage lines — only on YOUR lab / program scope.
2. Integrity paragraph
RoE + YOUR apps + authorized red team key → professional handoff idea from YOUR lab path.
Name what breaks if you skip scope or test stranger apps — ethics tie-in.
3. Shutdown
Notes secured chmod 600; index signed complete UTC.
Ready for Topic 33 Threat Intelligence on separate authorized track.
4. What you ship: capstone evidence pack complete
Checked checklist + sealed literacy pack + integrity paragraph + zero refused scan.
5. What you record before the next lesson
Capstone pack path. $AI_LAB quiet.
6. Wrong vs right: stranger SaaS vs YOUR toy LLM
Worked failure — same MSF word, opposite target. Right never needs a café Wi-Fi or classmate laptop.
- Wrong: Capstone includes jailbreak weaponization walkthrough as bonus exhibit.
- Right: Capstone complete on $AI_LAB with full teardown. Next: topic quiz.
Mission: Topic 32 capstone
1) Walk checklist with evidence links. 2) Integrity paragraph from your path. 3) Seal pack — zero refused files. 4) Sign index UTC complete.
Stuck? Ask Cyberlium AI Mentor
Ask Mentor: “Minimum N/A documentation?”
APPLY: Capstone target:
APPLY: True or False: Jailbreak weaponization earns capstone credit.
APPLY: Capstone teardown includes: