[Feature] Skill signing and integrity verification (ed25519 + content_hash)

[rohitg00]

Problem

The agent skills ecosystem is under active attack (Q1 2026):

• ClawHavoc campaign: 341 malicious skills flooding ClawHub in 3 days
• Snyk audit: 13.4% of marketplace skills contain critical security issues
• OWASP Agentic Skills Top 10 recommends ed25519 signing + content_hash

SkillKit has skillkit scan and skillkit validate but no cryptographic integrity verification.

Proposed Solution

Phase 1: Content Hashing

• Add integrity field to skill.json (per [Feature]: Read skill.json from source repos for install, recommend, and translate #83): "integrity": "sha256-yY1jg1cPGoisxK/..."
• skillkit install verifies hash after download
• skillkit validate checks integrity matches content
• skillkit publish auto-generates integrity hash

Phase 2: Skill Signing

• ed25519 key pair generation: skillkit sign init
• Sign skills on publish: skillkit publish --sign
• Verify signatures on install: skillkit install --verify
• Public key registry for known publishers

Phase 3: Trust Chain

• Configurable trust levels: trusted (signed + verified), scanned (passed security scan), unknown
• skillkit.yaml config: trust: { requireSigned: true, allowedPublishers: [...] }
• CI/CD flag: --require-signed to fail on unsigned skills

References

• OWASP Agentic Skills Top 10: https://owasp.org/www-project-agentic-skills-top-10/
• skill.json spec (integrity field): [Feature]: Read skill.json from source repos for install, recommend, and translate #83
• CVE-2025-59536 / CVE-2026-21852: Claude Code config injection via repository files