What problem does it solve? API keys and credentials stored in plaintext .env files, settings JSON env blocks, or shell history are exposed to backups, cloud sync, git commits, and any process with file read access. This Skill encrypts secrets at rest with FIPS 203 ML-KEM-768 and AES-256-GCM, injecting them into process memory only at runtime. ## Core Features & Use Cases - PQC-encrypted secrets bundle: A single per-machine bundle at ~/.config/pqc-secrets/secrets.bundle.json holds all API keys, encrypted via a double-envelope (ML-KEM-768 KEM, AES-256-GCM keywrap and data layers) with a passphrase-wrapped vault.pqc identity root using Argon2id and ML-DSA-65 signed audit chains. - Dual-engine CLI: A Rust binary (darwin/arm64 fast-path) and a canonical Python engine (pyca/cryptography>=45 native ML-KEM-768) provide keygen, pack, export, verify, list, rename, migrate, and vault commands with interoperable bundle JSON. - Agent and app integration: Ten browser_secrets_* MCP tools, an append-only SHA3-256-fingerprinted audit log, and integration patterns for Claude Code, VS Code, shell wrappers, CI/CD, and long-running applications like local-router. - Use Case: Wire Claude Code so settings.json contains only an empty ANTHROPIC_API_KEY placeholder, then run secrets-load before launching so the real key exists only in the OS keychain and process memory. ## Quick Start Ask the agent to generate a PQC keypair with pqc-secrets keygen, pack your API keys into the encrypted bundle, and load them into the shell with eval "$(pqc-secrets export)" before launching your tool.