Security & Testing Agent Skills for Claude Desktop

Harden an LLM application against sensitive-data leakage by filtering and redacting model output and constraining what reaches the context.

Treat LLM output as untrusted and encode it for the specific sink that consumes it — HTML/JS context encoding, parameterized queries…

Confirm and demonstrate direct and indirect prompt injection against an LLM-backed application during an authorized assessment.

Harden an LLM application's system prompt so it carries no secrets or authorization logic and cannot be trivially extracted.

Gate every document that enters an LLM application's knowledge base or fine-tuning corpus behind source allowlisting, provenance signing…

Demonstrate unbounded resource consumption against an LLM application — token floods, unbounded output, and cost/denial-of-wallet…

Enforce tenant and user isolation in a shared vector store by filtering retrieval on an access boundary before ranking, using…

Retrieve another tenant's or another user's documents from a shared vector store by exploiting similarity search that is not bounded by an…

Performs evidence-first, read-only static analysis on authorized local executables, application binaries, and release artifacts.

Read loop-constraints.md at the start of every run and enforce every rule.

Independent verification agent for loop-produced changes.

Independent verification agent for loop-produced changes — runs tests, confirms diff scope, and issues APPROVE / REJECT / ESCALATE_HUMAN.

Use after a failure has occurred (bug shipped, test broke, deploy rolled back, data lost, security incident) — the impulse is to fix the…

Performs evidence-driven root-cause diagnosis and regression-tested fixes for technical failures, build errors, and runtime bugs.

macOS app signing artifacts: inspect code signatures, entitlements, hardened runtime, sandbox, Gatekeeper, and trust failures; exclude…

macOS Xcode and SwiftPM tests: run focused scopes and diagnose build, assertion, crash, async-flake, fixture, entitlement, and host-app…

Periodic pass that keeps a project's verification skill and feature map honest: parallel source readers per feature, one live session…

Threat-model product features, APIs, data flows, secrets, permissions, supply-chain changes, auth boundaries, and risky code paths before…

Define behavioral contracts, run adversarial tests, and detect regressions for AI agents — invariants, edge cases, statistical analysis…

Spawn a fresh adversarial reviewer mid-implementation for any non-trivial decision — branching logic, cross-module boundaries…