origammi-digital/agents-marketplace
Agent
LLM evaluation engineer using DeepEval. Designs test suites that verify agent and skill behavior across runs — correctness, groundedness, context relevance, hallucination detection, and regression tracking. Activate when testing LLM outputs, building eval pipelines, catching prompt regressions, or validating agent…
origammi-digital/agents-marketplace
Skill Claude CodeCodex
LLM evaluation engineer using DeepEval — correctness, groundedness, hallucination detection, prompt injection defense, agent regression suites.
origammi-digital/agents-marketplace
Plugin Claude Code
Meta skills for maintaining the marketplace itself: how to author, edit, and test skills so they actually get used.
origammi-digital/agents-marketplace
Agent
Use at the start of any task to decide which marketplace skills apply and in what order — before writing code, exploring, or even asking clarifying questions. Establishes how to discover skills, the process-before-role priority, and the habit of announcing and checklisting a skill you invoke.
origammi-digital/agents-marketplace
Agent
Use when creating, editing, or reviewing a skill in this marketplace — authoring a new agent, rewriting an existing one, or deciding why a skill "isn't being used." Covers the two skill shapes (role vs process), the description-is-a-trigger rule, rationalization tables, positive-recipe wording, multi-platform action…
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Bootstrap skill: how to discover and sequence the marketplace skills at the start of a task — process-before-role priority, description-as-trigger discovery, announce-and-checklist. Injected into every session by the agents install-hook SessionStart hook.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
How to author, edit, and test a marketplace skill — role vs process shapes, description-as-trigger, rationalization tables, positive-recipe wording, multi-platform action language, and testing skills with llm-eval.
origammi-digital/agents-marketplace
Plugin Claude Code
Engineering process discipline: how the work is conducted, not who does it. Debugging, TDD, planning, and verification — the workflow skills that resist being skipped under pressure.
origammi-digital/agents-marketplace
Agent
Use when a task is multi-step, spans more than one file or system, or has any unknowns — before writing code. When you're tempted to "just start coding" a feature whose shape isn't fully clear. Turns a fuzzy request into a decomposed, reviewable plan of bite-sized steps.
origammi-digital/agents-marketplace
Agent
Use when facing any bug, test failure, crash, or "it works sometimes" — before proposing or trying a fix. Especially when you've already tried one fix that didn't work, or you're tempted to guess. Covers reproduce → isolate → root cause → fix → verify, and the trap of patching symptoms.
origammi-digital/agents-marketplace
Agent
Use when implementing any feature or bugfix, before writing implementation code. When you're tempted to "write it first and add tests after," or think the change is too small to test. Covers the red-green-refactor loop and why the test comes first, not after.
origammi-digital/agents-marketplace
Agent
Use before claiming any work is done, fixed, passing, or ready — before committing, opening a PR, or telling the user it works. When you're about to say "done" based on what the code should do rather than what you observed it do. Requires fresh evidence, not confidence.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Turn a fuzzy, multi-step request into a decomposed, reviewable plan of bite-sized steps before coding. Use for anything spanning more than one file or carrying unknowns.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Reproduce → isolate → root cause → fix → verify. Use before proposing any fix, especially after a fix already failed or when tempted to guess. Kills symptom-patching.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Red-green-refactor: the failing test comes before the code, for features and bugfixes alike. Use before writing implementation code.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Require fresh, just-observed evidence before claiming done/fixed/passing. Use before committing, opening a PR, or telling the user it works.
origammi-digital/agents-marketplace
Plugin Claude Code
Product and squad management: PM for discovery/prioritization, manager for orchestrating the full team.
origammi-digital/agents-marketplace
Agent
Squad orchestrator. Routes requests to the right agents in the right order. Enforces quality gates (security, QA, legal when relevant). Never implements code. Activate as the entry point for any development request — feature, bug, refactor, or security fix.
origammi-digital/agents-marketplace
Agent
Senior product manager. Translates user problems into scoped, prioritized work. Reads project context before opining. Identifies what NOT to build as much as what to build. Uses JTBD framing and Given/When/Then acceptance criteria. Activate for feature discovery, scope definition, prioritization, or friction analysis.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Squad orchestrator — routes requests, enforces quality gates, never implements code.
origammi-digital/agents-marketplace
Skill Claude CodeCodex
Senior product manager — JTBD discovery framework, Given/When/Then acceptance criteria, scope protection, friction analysis.
origammi-digital/agents-marketplace
Plugin Claude Code
On-demand reference skills: detailed code patterns invoked by other agents when needed. Install to make them available as context.
origammi-digital/agents-marketplace
Agent
Reference: concrete code patterns for N+1 detection, unbounded queries, sequential vs parallel async, and partial failure consistency. Invoke when reviewing or fixing performance issues. Companion to architect and dev-team-backend.
origammi-digital/agents-marketplace
Agent
Reference: test code patterns for PHPUnit (Laravel), Vitest + React Testing Library, Playwright (E2E), and Go (testify). Invoke when implementing or reviewing tests and you need concrete framework syntax. Companion to dev-team-tester.