Dadmin88/hermes-agency

accessibility-audit

Perform a structured accessibility audit of a product experience or implementation and produce prioritized remediation guidance.

accessible-design-review

Review product design artifacts before implementation for accessibility of information, interaction, focus, input, error recovery, responsive/reflow behavior, motion, contrast, content alternatives, and assistive-technology semantics.

keyboard-testing

Validate a user flow with keyboard or equivalent non-pointer input by checking reachability, order, focus visibility, activation, escape, shortcuts, custom widgets, and recovery from dynamic UI changes.

screen-reader-review

Review a critical interface flow with representative screen-reader and accessibility-tree evidence, checking names, roles, states, structure, navigation, announcements, reading order, and task completion.

wcag-review

Review a product flow or implementation against an explicitly identified WCAG version and conformance target by mapping applicable success criteria to reproducible evidence without reducing conformance to scanner output.

agent-failure-analysis

Diagnose AI and agent failures from trace evidence by locating the first meaningful divergence across routing, instructions, context, retrieval, tools, state, model behavior, schemas, and runtime.

llm-evaluation

Design and run an evaluation for an AI or agent feature using representative tasks, explicit success criteria, failure taxonomy, baselines, and reproducible evidence.

model-selection

Select an AI model for a real workload using representative evaluations, current official capability and pricing data, latency and cost measurements, operational constraints, and explicit fallback tradeoffs.

prompt-design

Design, version, and improve prompts as testable behavior contracts with explicit instructions, trustworthy context boundaries, examples, output requirements, and evaluation evidence.

retrieval-design

Design retrieval for AI features around corpus quality, permissions, chunking, indexing, query strategy, ranking, context assembly, provenance, and retrieval-specific evaluation.

structured-output

Design and validate model outputs as explicit machine-readable contracts with typed schemas, semantic checks, bounded recovery, versioning, and failure handling.

tool-use-design

Design model-facing tools with narrow capability boundaries, clear schemas and descriptions, safe side effects, explicit failure semantics, observability, and task-level evaluation.

attribution-analysis

Analyze marketing attribution using explicit touchpoint scope, identity rules, conversion windows, model assumptions, incrementality limits, and cross-channel evidence without treating attributed credit as causal proof.

cohort-analysis

Analyze behavior over time by cohort using consistent entry events, maturity windows, retention or value measures, and segment context so lifecycle patterns are not distorted by mixing users at different ages.

dashboard-spec

Specify an analytics dashboard around decisions, metric definitions, dimensions, freshness, ownership, thresholds, drilldowns, data quality, and interpretation rather than filling a screen with available charts.

experiment-analysis

Analyze a marketing or product experiment from assignment integrity, exposure, primary and guardrail metrics, uncertainty, segment effects, novelty, duration, and business relevance.

funnel-analysis

Analyze a marketing or product funnel from qualified entry through successive outcomes using trustworthy event definitions, segment comparison, leakage diagnosis, and decision-relevant conversion evidence.

measurement-plan

Design a measurement plan that links business or product questions to metrics, events, dimensions, data quality, and decision thresholds.

art-direction

Establish and review a visual art direction with clear principles, references, constraints, quality bars, and consistency across assets.

asset-quality-bar

Define and apply a measurable acceptance bar for visual assets covering artistic coherence, readability, technical fitness, reuse, and final-context validation.

reference-brief

Build a focused visual reference brief that translates creative intent into usable examples, attributes, anti-references, and production guidance without encouraging imitation.

style-guide

Codify an approved visual language into durable rules, examples, ranges, and exceptions that keep distributed production coherent without freezing creative judgment.

visual-review

Review visual work against approved art direction, hierarchy, composition, readability, consistency, production constraints, and intended emotional effect.

3d-optimization

Optimize 3D assets for runtime while preserving silhouette and material intent through geometry, LOD, UV, draw-call, skinning, collision, and import decisions.

asset-production

Plan and produce a 2D or 3D asset from brief through references, construction, optimization, export, validation, and handoff.

export-validation

Validate exported art assets for format, scale, orientation, pivots, naming, channels, transforms, animation, metadata, and target-tool import behavior before handoff.

sprite-production

Produce sprites and sprite sheets with clear silhouettes, consistent scale, animation-ready registration, clean transparency, sampling discipline, and runtime validation.

texture-workflow

Create and validate texture sets with intentional material definition, texel density, channel packing, tiling, color management, compression, and runtime context.

ambience-design

Design layered environmental ambience that communicates place, state, time, and activity through beds, emitters, variation, transitions, and silence without becoming repetitive or masking important cues.

audio-design

Design an audio treatment for an interactive experience using functional cues, emotional tone, layering, variation, spatial context, and implementation constraints.

audio-implementation-plan

Plan how music, ambience, effects, dialogue, UI cues, spatialization, states, priorities, and runtime controls integrate with product or game behavior.

loudness-mix-review

Review an interactive or linear mix for intelligibility, balance, dynamic range, loudness, frequency masking, user controls, platform translation, and state consistency.

music-brief

Create a composer-ready music brief defining emotional arc, narrative or gameplay function, structure, adaptive states, instrumentation direction, references, deliverables, and implementation constraints.

automation-design

Design a reliable automation from trigger through validation, idempotency, side effects, retries, observability, failure handling, and human escalation.

idempotent-automation

Make automation safe under duplicate triggers, retries, restarts, overlapping runs, and uncertain outcomes by defining stable operation identity, state checks, atomic effects, and reconciliation.

job-observability

Observe automated jobs through trigger, logical operation identity, queue/wait time, attempts, execution, side effects, checkpoints, failures, completion, and stale/stuck detection.

retry-design

Design retries around failure classification, safe repetition, bounded attempts, backoff and jitter, deadlines, capacity, circuit breaking, reconciliation, and escalation instead of retrying every error blindly.

workflow-automation

Implement multi-step workflow automation with explicit state, prerequisites, branching, side-effect boundaries, checkpoints, resumability, approvals, cancellation, and final reconciliation.

api-design

Design or evolve a backend API with explicit contracts, protocol semantics, compatibility, authorization boundaries, failure behavior, and consumer validation.

auth-integration

Integrate an accepted authentication and authorization design into backend services without weakening trust boundaries, credential handling, or access-control guarantees.