@mindbomber

aana-purchase-booking-guardrail enforces verification gates across item, price, terms, authorization, payment privacy, and reversibility before any financial commitment, with explicit approval rules and safer alternatives to reduce downstream fraud or user regret.

aana-file-operation-guardrail enforces scoped, reversible, and authorized file operations by requiring explicit scope verification, pre-flight checks, and human approval before destructive or cross-boundary actions on user files.

aana-workflow-readiness-check is a pre-execution gatekeeping skill that classifies workflows into seven readiness states (ready, needs-information, needs-permission, needs-evidence, needs-tool-access, needs-review, not-safe) and applies structured decision rules to information, permission, tools, evidence, and risk before starting multi-step tasks.

guardrail skill for customer support replies that enforces fact verification, policy compliance, private data minimization, and prevents unauthorized promises. provides decision framework and pre-flight checks for support agents.

aana-meeting-summary-checker validates meeting notes for grounding in evidence, clear ownership, confirmed dates, privacy safety, and appropriate attribution before sharing or applying summaries.

guardrail skill for email composition and send validation. enforces pre-send checks on recipients, privacy, attachments, claims, tone, and approval status before any send or schedule action. instruction-only, no direct service integration.

aana-task-scope-guardrail provides a structured loop and decision framework for agents to identify task boundaries, classify proposed actions, and stop when complete without expanding beyond user intent or accessing unrelated data.

aana-evidence-first-answering provides a structured framework for separating known facts, assumptions, inferences, and missing evidence before committing to answers or actions. it enforces caution in high-risk domains by requiring explicit classification and retrieval decisions before conclusions are finalized.

aana-publication-check provides a pre-publication gate loop and checklist for classifying content readiness across authorization, accuracy, privacy, permissions, scope, and quality. instruction-only, designed for agent-guided review rather than automated enforcement.

aana-medical-safety-router establishes guardrails for health-related agent responses by classifying queries, detecting emergency signs, preventing diagnosis and treatment overclaims, and routing to appropriate care. instruction-only, no external integrations.

aana-continuous-improvement provides a bounded loop for agents to identify workflow failures, propose localized fixes, and verify changes without modifying instructions or permissions. focused on low-risk iteration within task scope.

aana-guardrail-bundled wraps a local python helper that forwards minimal policy review payloads to a trusted localhost aana bridge, with explicit safety boundaries and decision rules for governance workflows.

aana-guardrail enforces safety checkpoints before agent actions that risk compliance violations, data exposure, or irreversible changes by routing requests through predefined decision gates and adapter rules.

Ensures research answers and reports strictly cite and align with allowed sources, clearly marking unsupported or uncertain claims before publishing.

Ensures private data is used only when necessary, authorized, minimized, and safe for tasks involving sensitive account, billing, health, legal, or personal...

Reviews and verifies code changes for scope, correctness, security, testing, secrets, and risks before commits, PRs, or deployments.

Creates a compact, privacy-aware audit record for important agent decisions, checks, changes, evidence, and residual risks.

Routes and classifies personal finance questions to provide educational responses, disclose risks, avoid unsupported claims, and refer high-impact issues to...

Routes, classifies, and reviews legal-related questions to provide general info, enforce jurisdiction caveats, avoid unauthorized advice, and refer high-risk...

Controls and reviews external tool use for necessary, scoped, authorized, data-minimized, and safe operations that may affect state or reveal private data.

Routes uncertain, high-impact, or sensitive actions to appropriate human or professional review before proceeding.

Manages user memory operations by validating relevance, sensitivity, source, approval, and lifecycle before storing, reusing, editing, importing, exporting,...

Ensures calendar event changes meet attendee, time, privacy, and approval criteria before allowing creation, updates, cancellations, invites, or shares.

Checks and enforces narrow, authorized, privacy-safe controls on data exports before file, record, or dataset sharing to prevent unauthorized or excessive da...

Validates message destination, audience, tone, privacy, and approval before sending or posting on Slack, Teams, Discord, SMS, or public channels.

Checks release readiness by verifying artifacts, tests, docs, approval, risks, and evidence before publishing tags, packages, or deployment notes.

Checks ticket updates for accuracy, evidence, privacy, authorization, and risk before allowing changes to customer-visible or internal support records.