Your agent reads user-pasted text, web pages, retrieved docs, or tool output — and you can't tell which sentence an attacker wrote. A one-line "ignore malicious instructions" never held. This delivers a layered injection-defense design: threat model, attack-surface map, input/privilege/output defenses, canary tokens and a classifier, an adversarial red-team set with before/after, and residual risk plus an incident runbook.
Turn a long recurring agent prompt into a short pointer prompt plus a procedure file, with a keep/delete table for every sentence.
Decide whether to buy, wait, or skip before you spend — with regret-risk scoring, future-self checks, and a 7-day purchase test.
I built this Skill to help users protect API keys, OAuth tokens and MCP credentials when working with AI agents in Codex Desktop, Claude Desktop and local automation environments. It combines static leak detection, vault preflight checks, runtime redaction, MCP-safe remediation, credential-store guidance and security reporting. The Skill is designed to reduce accidental secret exposure while preserving legitimate MCP and API workflows.
Checks timesheets against your overtime and break policy, flags issues before you run payroll. Works from your own stated rules, not guesses.