All notable changes to CyberStrike are documented in this file.
Format based on Keep a Changelog, versions follow Semantic Versioning.
1.1.16 — 2026-08-08
- linuxhook — 120 native TypeScript post-exploitation programs across 9 categories (recon, credentials, privesc, persistence, lateral movement, evasion, exfiltration, network attacks, IPv6) with bash/sh/python3/perl fallback chain
- llmhook — 23 LLM security testing programs (OWASP LLM Top 10), 130+ probes/payloads with
--format openai|anthropic|generic - containerhook — 7 Docker container security programs (escape, privesc, pivoting)
- iachook — 8 Terraform/IaC audit programs
- k8s-audit — 10 read-only Kubernetes security assessment programs
- ci-audit — 8 CI/CD pipeline security programs (GitHub Actions, GitLab CI, Jenkins, CircleCI)
- gcphook — native GCP post-exploitation tool
- cloud_audit — native cloud security audit tool for
internal-networkandcloud-securityagents - Skills:
linux-postexploit(kill chain + MITRE ATT&CK),ci-assessment - cve-mcp dynamic CTA — all OS hooks guide agents to query
cve-mcpfor discovered services
- winhook — modular refactor (16 files, 154 programs) + cmd.exe/reg.exe/wmic fallback
- machook — modular refactor (10 files, 46 programs) with new credential/lateral/evasion handlers
- azurehook, kubehook, cipipe — converted from Python to native TypeScript
macos-postexploitskill — updated from 12 to 46 programs
- Hackbrowser: shell quoting error in Chromium path detection
- Hackbrowser: Esc now stops the crawl subprocess (previously continued in background)
- Build: bundled migration crash on first-time setup (Drizzle ORM 1.0 missing
namefield) - Provider: Vertex AI autoload, regional endpoints, Copilot stale response IDs, Gemini/Copilot schema sanitization, overflow detection for 5 new providers, Devstral Turkish locale fix, Codex compatibility
- MCP: orphan process cleanup, token masking, tool execution crash handling, URL validation
- Session: auto-compaction toggle, content-filter surfacing, empty message filtering, reasoning fallback
- Config: CRLF frontmatter, invalid permission JSON, inaccessible config dirs
- Tools: destructive edit prevention, symlink grep, script binary detection, Windows permission paths
- Other: entity double-escaping, ID generation modulo bias, backslash escaping in sanitization
- fast-xml-parser 5.9.3 → 5.10.1 (DoS via XML parsing)
- dompurify 3.4.11 → 3.4.13 (XSS via hook removal)
- inject-probe — 13 new test files (payload battery, cross-location injection, WAF detection, XSS routing)
- vuln-scope — 7 new test files, 166+ test cases
1.1.15 — 2026-06-23
- eBPF post-exploitation tool — 10 kernel-level programs for credential harvesting, process/file/connection hiding, system monitoring, and cleanup
- eBPF blind spot monitors — 20 kernel-level detection programs covering io_uring bypass, memfd fileless exec, ptrace injection, VDSO side-channels, namespace escape, and 14 more attack primitives
- winhook — 12 Windows post-exploitation programs (AV/EDR evasion, credential harvesting, monitoring, cleanup)
- machook — 12 macOS post-exploitation programs (Keychain, Chrome creds, SSH keys, TCC bypass, DTrace tracing)
- awshook — 10 AWS post-exploitation programs (IAM privesc, S3/Secrets dump, Lambda backdoor, CloudTrail evasion)
- azurehook — 8 Azure post-exploitation programs (Entra ID, Key Vault, managed identity, runbook backdoor)
- kubehook — 7 Kubernetes post-exploitation programs (secrets extraction, container escape, RBAC privesc)
- cipipe — 5 CI/CD attack programs (GitHub Actions/GitLab/Jenkins secret extraction, pipeline injection)
- Skills:
ebpf-attacks,windows-postexploit,macos-postexploit,aws-postexploit,azure-postexploit,k8s-postexploit,cicd-attacks— all with MITRE ATT&CK mappings - GitHub Copilot Enterprise provider — Claude/GPT/Gemini at zero cost through Copilot, OAuth device flow auth
- DAST proxy-testing memory — per-credential coverage notes, observed values for IDOR substrate
- GraphQL & JSON-RPC as first-class endpoints with deterministic op-keys
web_get_detailtool, vulnerability triage lifecycle with grouped views, hybrid HackerOne-ready report generation
- Bounded session context —
web_get_session_contextreturns scoped data instead of full session dump (O(N²) → O(1) per call)
- Endpoint over-fragmentation for form/multipart POST bodies
- Phantom hosts from
//-relative targets and HTTP/2:authority - Silent vulnerability loss in dedup path (~50% drop rate → 0)
- Proxy-flow ordering (analyzer blocks before tester subagents)
- Hackbrowser crawl coverage (shadow-DOM, web components, framework bindings, ephemeral ID rejection)
- Hackbrowser provider routing for non-OpenAI keys
- Token cost double-charging for reasoning tokens
1.1.6 — 2026-03-30
- Web UI v1.1.6-beta — branding, auth, side panel, offensive tooling
- Web UI bundled in npm package (auto-installs to
~/.cyberstrike/web/) - MCP/Bolt status tabs in TUI status popover with config persistence
- CORS and auth failures on remote/tunnel access
- Enterprise infra made conditional on
CYBERSTRIKE_ENTERPRISEenv var - Stripe/PlanetScale providers made optional
1.1.4 — 2026-03-18
- Intelligence layer, SEO optimization, Bolt 1:N architecture in README
--betaflag for install script
- Schema reconciler for partially applied migrations
- Auto-fallback to available port when default port busy
- npm scope renamed
@cyberstrikeus→@cyberstrike-io - All
cyberstrike.us→cyberstrike.io - Bin launcher
opencode→cyberstrike
- License consolidated to AGPL-3.0-only
- Docker and Tauri desktop app removed (opencode legacy)
- 13 unused inherited workflows removed
1.1.0 — 2026-03-17
- Bolt remote tool server — Ed25519 key pairing, SDK fetch chain, live sidebar status, TUI management
- MCP server management — add/remove from TUI with validation
- Local LLM provider support — any OpenAI-compatible endpoint (vLLM, Ollama, LM Studio)
- 23-language README with social preview SVGs
- GitHub issue/PR templates, SECURITY.md, CONTRIBUTING.md
- Config discovery, lazy tool registry, Bus.subscribe timing, Bolt auth, opentui borders
1.0.8-beta.1 — 2026-03-15
- 13+ specialized security agents (web, mobile, cloud, internal-network, 8 proxy testing agents)
- 120+ OWASP WSTG test cases built into agent methodology
- Vulnerability reporting dialog and web proxy context tools
0.1.0 — 2026-02-16
- Initial public release of CyberStrike
- AI-powered offensive security agent platform with 13+ specialized agents
- Claude Code CLI/API provider, cloud security agent, chunked context compaction
- MCP browser server, ASCII logo