AI Agent Workflow Collection

DEFAULT for authoring or refining an Ultraprompt agent — produces a new agent definition with frontmatter, body, tool permissions, and orchestration plan. Different from /skill-author (authors a skill, not an agent), /plugin-review (audits an existing agent, not authors), /hooks-design (designs a hook, not an agent). Triggers: 'new agent, author an agent, refine this agent, agent definition, design a subagent, add a specialist'.

DEFAULT for AI agent safety reviews — dispatches security-auditor + risk-and-controls-reviewer with AI/agent safety focus.

DEFAULT for ROUTING AMBIGUITY — interactive picker that surfaces top skill candidates + 2 LLM-rewritten prompt variants via AskUserQuestion with previews, then dispatches the chosen skill with the chosen phrasing: AskUserQuestion-driven selection moment when the V8 router can't decide; on choice, dispatches the picked skill (or surfaces the slash command for manualonly skills) and logs the choice for the learning queue. Different from /route (read-only routing table — this is interactive picker), /menu (catalog browser — this is scoped to the current prompt), /pathfind (workflow path across multiple skills — this picks ONE skill). Triggers: 'auto-invoked by the V8.7 UserPromptSubmit hook when top-2 skills are within 15% or top is medium confidence; manually invokable as /ultraprompt:choose '.

DEFAULT for BUILD/LINT/TYPECHECK/PIPELINE FAILURE — fix the pipeline, not the symptom: layered failure analysis (workflow|cache|deps|env|code), last-successful-run diff, fix, and green-run command. Different from /debug (runtime/test failures in code), /dependency-audit (CVE-driven, not failure-driven), /release (release notes, not pipeline repair). Triggers: 'CI is red, pipeline broken, build is failing, lint failure, typecheck failure, cache miss'.

DEFAULT for concept-stage thinking — produces a structured concept brief (problem, target users, proposed approach, key differentiator, success criteria, top risks, validation plan, decision required). Different from /prd-lite (early-stage PRD), /idea-triage (multi-idea ranking), /problem-framing (problem definition only).

DEFAULT for contract-test generation — dispatches test-strategist + test-harden with contract focus.

DEFAULT for cost analysis spanning LLM tokens, cloud spend, and database query cost — produces a ranked findings list with monthly $-cost estimate, severity, and remediation suggestion per item. Different from /auditor (general single-concern auditor — this skill is specifically cost-focused with $-amount evidence), /performance-pass (latency, not $-spend), /supply-chain-hardening (dependency security, not cost). Triggers: 'cost audit, llm cost, cloud cost, expensive query, n+1, runaway tokens, save money, hot path spending'.

DEFAULT for privacy data-flow mapping — dispatches auditor with privacy + data-flow focus.

DEFAULT for UNUSED, STALE, OR DRIFTED CODE ONLY — exports/imports/configs/migrations/fixtures with no live reference: ranked drift findings (safe-to-remove vs needs-review) with reference evidence and quick wins. Different from /gap-analysis (missing features, not unused code), /feature-completeness (single-feature completeness), /technical-debt-triage (debt with risk×leverage scoring, not drift), /repo-review (whole-repo audit). Triggers: 'dead code, unused exports, stale fixtures, drift, what can we delete'.

DEFAULT for AUDITOR-ONLY single-feature completeness check (when you already know which feature and do not want orchestration): structured completeness audit for one feature with confirmed/likely/missing controls and file:line evidence; no cross-feature orchestration. Different from /gap-analysis (orchestrates this auditor + wiring-gap-inspector — use that for cross-cutting feature gaps), /repo-review (whole repo), /test-gap-analysis (missing tests only). Triggers: 'is complete, audit for completeness, does handle '.

DEFAULT for plugin hook design — dispatches reviewer/architect with plugin hooks focus.

DEFAULT for deciding what to build from a known set of options: runs structured triage of existing ideas with weighted criteria (customer impact, strategic fit, effort, evidence strength, risk). Different from /innovation-lead (idea generation), /concept-brief (single concept polish), /opportunity-map (landscape framing, not triage).

DEFAULT for IaC reviews — dispatches reviewer with infrastructure-as-code focus.

DEFAULT for LLM/agent eval design — dispatches evaluator for AI/LLM-specific evaluation design (offline + online metrics, groundedness, hallucination, drift, cost, latency).

DEFAULT for MCP design tasks — dispatches reviewer/architect with MCP focus.

DEFAULT for observability-specific audits — dispatches auditor with observability focus.

DEFAULT for authoring a new-engineer onboarding doc — produces a structured onboarding guide with setup steps, codebase tour, key concepts, common workflows, and a 'first PR' exercise. Different from /docs-sync (documentation drift correction, not new authoring), /repo-map (machine-readable repo map, not human-facing onboarding), /runbook-author (operational, not onboarding). Triggers: 'onboarding doc, new engineer setup, first day at the repo, codebase tour, getting started, dev environment setup'.

DEFAULT for opportunity space exploration: produces structured opportunity map with market/competitive/customer/internal axes, sized opportunities, evidence per zone, and recommended focus: runs the opportunity-map discipline. Different from /idea-triage (ranks existing ideas) and /concept-brief (drafts one chosen concept). Triggers: 'opportunity map / what opportunities exist / market opportunity analysis / where are the openings / strategic opportunity assessment / opportunity space / what should we explore'.

DEFAULT for performance audits: dispatches reviewer with performance focus: runs the performance-pass discipline. Different from /debug (active perf bug). Triggers: 'performance audit / find slow code / hot path review / perf optimization / performance issues / latency review / throughput audit'.

DEFAULT for PRD FOR AN AI/LLM/AGENT FEATURE — model selection rationale, eval plan, safety boundaries, cost envelope, failure-mode map: PRD covering AI-specific sections (model selection, eval design, safety boundaries, cost envelope, failure-mode map) on top of the standard PRD shape. Different from /prd-standard (non-AI features), /prd-technical (deep tech-design PRD, not AI-specific), /llm-eval-design (eval design only). Triggers: 'PRD for an AI feature, AI-feature spec, agent feature PRD, LLM feature spec, model-backed feature'.

DEFAULT for early-stage thinking that doesn't yet warrant a full PRD — produces a 1-2 page structured product doc with problem, users, goals, scope, success criteria. Different from /prd-standard (full PRD with users + scope + risks), /prd-technical (infra-heavy PRD), /concept-brief (single concept, not yet a product).

DEFAULT for any product change crossing teams or with meaningful complexity — produces a full structured product doc with problem, users, goals, non-goals, must/should/won't-have, scope, technical considerations, risks, metrics, acceptance criteria, rollout plan, validation plan, open questions. Different from /prd-lite (1-2 page early-stage), /prd-technical (infra-heavy), /prd-ai-feature (LLM/agentic feature).

DEFAULT for infrastructure work, platform features, developer tools, or any product where technical shape is the primary risk — produces a PRD oriented around technical design with deeper technicalconsiderations, data model implications, API surface, integrations, failure modes, telemetry. Different from /prd-standard (customer-journey-heavy), /prd-lite (early-stage), /architect (code-level architecture, not product spec).

DEFAULT for WHOLE-REPO end-to-end audit (map + gaps + drift + test gaps + release readiness in one pass): structured repo review covering map, confirmed gaps, probable gaps, test gaps, contract drift, stale code, release readiness, top risks, quick wins, and implementation sequence — V8 panel-ready. Different from /gap-analysis (ONE feature end-to-end), /feature-completeness (single feature), /dead-code-drift (drift only), /test-gap-analysis (tests only), /release-readiness (ship/no-ship gate only). Triggers: 'audit the codebase, review the whole repo, what's incomplete across the repo, is this ready to ship overall, comprehensive repo audit'.