The work
All 30 authored projects
Everything below is authored work published under the MIT licence. Test counts
are the number each repository's own suite reports, measured by running it on a
fresh clone. Star counts come from the live GitHub API at build time.
All
Red-team and detection
AI and agent security
Assessment and platform
Upstream contribution
Other
Loading projects
Full project list
JavaScript is off, so here is the same list as plain text.
Red-team and detection engineering
accessweave — Offline identity-relationship and directory-permission review from explicitly exported inventory data. 33 tests.
cipherbench — A benchmark and interoperability test harness specification for vetted post-quantum cryptographic providers. 29 tests.
fixqueue — Merge SARIF, Trivy, Grype and OSV scanner output into one deduplicated, EPSS and KEV prioritized fix queue. 129 tests.
fuzzharbor — A reproducible local fuzzing campaign manager and crash triage toolkit. 21 tests.
iamosiac — Offline cloud-IAM policy review and permission-relationship visualization data, with least-privilege remediation hints. 39 tests.
opsledger — Tamper-evident evidence ledger for red-team engagements: hash-chained events and a signed manifest. 115 tests.
quanta-sentry — Cryptographic inventory and post-quantum migration-readiness reporting from source and configuration evidence. 26 tests.
rulebench — Sigma rule conformance harness: run rules against events, lint them, and measure coverage against the MITRE knowledge base. 187 tests.
secret-sniper — Leaked credential detection with live verification. 69 selftest checks.
violet-trace — A safe purple-team validation harness for defensive telemetry and detection-rule tests. 34 tests.
AI and agent security
context-bridge — A VS Code extension that writes privacy-conscious Markdown handoffs, with secret redaction built in. 66 tests.
gravion-harness — A scoped, auditable security-testing harness with a hash-chained audit log. 127 tests.
gravion-trainner — Model quality evaluation starter for OpenAI-compatible chat-completions endpoints. 27 tests.
llm-security-bot — Pattern-based static analysis for code, configs, log lines and LLM input — deterministic, so the output is a triage list, not proof of exploitability. 20 tests.
local-slm-eval-kit — Evaluates a local model against a fixed task suite with redaction, timeouts and diffable reports. 108 tests.
memorytrust — Audits an agent memory store for unverified, stale, contradictory and sensitive content and scores how much can be trusted. 112 tests.
preference-data-doctor — Diagnoses and repairs preference datasets: duplicates, contradictions, class imbalance, format drift, leaked secrets. 147 tests.
skill-supply-chain-auditor — Audits agent skill packages for prompt injection, unsafe shell, obfuscated payloads, embedded secrets, unpinned dependencies. 125 tests.
Assessment, automation and platform
TRISHUL — Authorized-scope, AI-powered security testing platform with real tool integration behind scope gates and an AI policy boundary. 118 tests.
aegis — Multi-agent system core: task lifecycle, persistent memory, provider abstraction, default-deny tool registry, budgets and recovery. 294 tests.
agent-browser — Visible Playwright browser starter with an explicit domain allow-list and explicit stop conditions. 43 tests.
api-security-testing — Read-only BOLA and cross-tenant detector with a rate-limit probe, backed by a local mock API for reproducible demos. 18 tests.
localshare — Shares a local development server through Cloudflare tunnels, with loopback-only origins. 40 tests.
privacyflow — Validates authorized WireGuard profiles and plans safe VPN rotation workflows. 38 tests.
rasgulla-ai — A local-first AI-assisted security research workbench for code understanding and safe test generation. 50 tests.
security-automation-scripts — Small Node.js and Bash utilities: header analysis, TLS configuration check, log parsing helpers. 22 tests.
web-security-labs — Four OWASP-style labs (IDOR, XSS, broken-auth, logic-flaw), each with a vulnerable and a secure variant, loopback-only by design. 22 checks across 4 labs.
webapp-security-toolkit — Authorized-only web and API security assessment toolkit: header analysis, TLS posture, scope enforcement, structured findings. 44 tests.
Upstream contribution
Other
catalog-forge — Crawls a shop catalog and serves it as a storefront.
Not listed here on purpose. The GitHub profile repository, the
portfolio website repository and a fork of another maintainer's project are not
authored projects, so they are not presented as work of this account.
Want one of these built for your stack, or maintained for your team?
See the services or
sponsor the tools .