AI Agent Workflow Collection

Add new stories to existing epics or create new epics with stories. Enforces story protection (in-progress/review/done stories are read-only), auto-increments IDs, and validates via inline Val integration. Use when "add stories".

Deep requirements elicitation using structured questioning techniques. Use when the user wants to explore requirements gaps, validate assumptions, and discover unstated needs using methods like 5 Whys, Socratic Method, User Story Mapping, MoSCoW, Kano Model, Jobs-to-be-Done, Assumption Mapping, and Stakeholder Mapping.

Universal review checklist, SOLID violation detection, cyclomatic and cognitive complexity thresholds, and the review-gate-completion hard gate enforced before a story moves to done. Shared dev skill JIT-loaded by dev-story, code-review, and stack dev agents.

Scaffold or regenerate a CI pipeline with quality checks. Use when "setup CI pipeline" or /gaia-config-ci (formerly /gaia-ci-setup); pass --regenerate to refresh generated workflows with the backup-before-overwrite UX and .user-steps.yml include pattern.

Display project-config.yaml read-only — top-level section TOC (default), a single named section, or the full byte-verbatim file (--full). Use when "show config" or /gaia-config-show.

Guide a human-centered design session through the five-phase Stanford d.school pipeline — Empathize, Define, Ideate, Prototype, Test. Use when "run design thinking" or /gaia-design-thinking. Delegates facilitation to Lyra (design-thinking-coach) and produces a creative artifact at .gaia/artifacts/creative-artifacts/design-thinking-{date}.md.

Edit an existing UX design document with cascade-aware downstream artifact detection, delegating UX-authoring reasoning to the ux-designer subagent (Christy) — planning skill. Use when the user wants to modify sections of an existing UX design while preserving consistency with architecture, epics, stories, and test plans.

One-gesture external-document ingestion into the Brain knowledge layer. Fetches a URL, local file, or stdin paste, auto-infers slug and tags, writes provenance-stamped markdown under the knowledge store, and registers the entry in the brain index.

Context-sensitive help with project-state-aware routing (greenfield / brownfield / post-update / healthy). Analyzes the user's query and current project state (which docs/ artifacts exist + stale-flag markers) to suggest the most relevant GAIA slash command. Primary intent-to-command map is ${CLAUDEPLUGINROOT}/knowledge/gaia-help.csv; every suggestion is cross-checked against ${CLAUDEPLUGINROOT}/knowledge/workflow-manifest.csv so the skill never invents command names. Use when "help" or /gaia-help, when the framework detects configuration drift (.framework-version-stale or .config-stale markers present), or when the user asks "what should I do" / "project state".

Design infrastructure topology and IaC structure through collaborative discovery with the devops subagent (Soren) — architecture skill. Use when the user wants to produce an infrastructure design document covering deployment topology, environment design, IaC structure, and observability plan.

Peer-to-peer multi-agent discussion skill — seven-phase lifecycle (INVITE / CHARTER / RESEARCH / DISCUSS / CLOSE / REVIEW / SAVE) with charter requirement, decide default mode, round-robin turn arbitration, and live-streamed transcript. Use when "/gaia-meeting" or "run a peer-to-peer meeting".

Automate the upgrade from GAIA v1 (workflow.xml engine) to v2 (Claude Code native plugin) — backup, migrate templates/memory/config, validate — and v2-to-v2 config reconciliation when an existing v2 project needs schema alignment after a framework update. Use after the v2 plugins have been installed via /plugin marketplace add.

Assess non-functional requirements covering performance, scalability, reliability, and security. Use when "assess NFRs" or /gaia-nfr.

Multi-agent group discussion with dynamic participant selection. Use when "party mode".

Run anytime performance bottleneck analysis on a story — N+1 queries, memory/bundle impact, caching, and algorithmic complexity. Emits a machine-readable PASSED/FAILED verdict and updates the Review Gate. Use when "performance review" or /gaia-perf-deepdive (formerly /gaia-performance-review).

Execute a project-generic release procedure — version bump (reading release.versionfiles[] from config), commit, tag, push, and GitHub Release. Wraps skills/gaia-release/scripts/version-bump.js and the main-only release branch policy. Use when "cut a release", "bump the version", or /gaia-release.

Plan a sprint by selecting stories from the backlog, applying sizing and priority rules via the sm subagent (Nate), and committing the sprint atomically to sprint-status.yaml via sprint-state.sh. GAIA-native replacement for the legacy XML engine workflow.

Craft a compelling narrative using story frameworks and emotional design. Use when the user asks to "craft a story", write a narrative, or build emotional copy around a product/idea. Delegates narrative construction to Elara (storyteller), picks a framework from the story-types catalog, and produces a ranked, polished narrative artifact.

Research a technology or tech stack with objective trade-off analysis. Use when the user wants to evaluate technologies, compare alternatives, and get adoption recommendations before architecture decisions.

Expand automated test coverage for a story. Use when "automate tests" or /gaia-test-automate.

Scan in-progress and completed story files for development findings and triage each into a new backlog story, an existing story, or dismiss. Produces new story files with complete frontmatter (15 required fields, status: backlog, sprintid: null). Source story findings tables stay intact for idempotent re-triage. Done-story guard blocks ADD TO EXISTING mutations against status: done targets with an explicit override path recorded for retrospective review. GAIA-native replacement for the legacy triage-findings XML engine workflow.

Remove an ingested document from the Brain knowledge layer. Deletes the ingested file and de-registers its brain-index entry atomically.

Full story validation with factual verification via Val subagent. Dispatches the validator via the main-turn Agent tool and records the outcome via review-gate.sh using canonical PASSED/FAILED/UNVERIFIED vocabulary.