Authors Guild plugin slash commands (commands/.md). Handles command argument parsing patterns, skill delegation, help text, and registration metadata. TRIGGER when a new /guild or /guild: slash command is needed, when an existing command's arguments/help need updating, or when a command must be re-wired to a new…
Authors and reconciles Guild plugin in-repo documentation (README.md, CONTRIBUTING.md, CHANGELOG.md, AGENTS.md/CLAUDE.md prose sections) and updates website-pointing links when the canonical docs site changes. The plugin no longer carries a docs/ set — docs/ holds only a static redirect page + assets. The primary…
Authors cross-cutting Guild plugin tests under tests/ — trigger-accuracy evals, wiki-lint fixtures, end-to-end regression specs, paired-eval harness tests, boundary-collision evals. TRIGGER when cross-cutting test fixtures or eval harnesses are needed, when regression coverage must be added at a phase gate, or when…
Authors Guild plugin hooks. Owns hooks/hooks.json plus hook scripts — bootstrap.sh, check-skill-coverage.sh, capture-telemetry.ts, maybe-reflect.ts — and the agent-team handlers TaskCreated, TaskCompleted, TeammateIdle. TRIGGER when a new Claude Code hook event needs wiring, when a hook script needs to be written or…
Lays down Guild plugin scaffolding (.claude-plugin/plugin.json, marketplace.json), repo-root CLAUDE.md, and top-level directory structure per the canonical repo layout. Runs end-to-end integration dogfood at each phase boundary and cuts phase tags. TRIGGER when starting a new plan phase, setting up plugin manifests…
Dev-team specialist for harvesting durable knowledge from Guild's own research/ideation corpus (.guild/research/ and the workspace self-build knowledge base) into recallable canonical pages. Distinct from the shipping guild:researcher (which writes end-user research deliverables) — this agent owns the SELF-BUILD path…
Authors Guild plugin skills across Tiers 1-5. Writes skill bodies, YAML frontmatter (name, description, whentouse), per-skill evals.json, and runs description optimization so every skill stays ≤ 1024 chars with ≥ 3 trigger phrasings. TRIGGER when a new skill is needed under skills/core/, skills/meta/…
Authors the Guild specialist role definitions — the 2 machinery agents under agents/.md and the 15 domain type templates under templates/specialists/.md. Writes pushy TRIGGER / DO NOT TRIGGER blocks, frontmatter (name, description, model, tools, skills), and role body guidance. Runs adjacent-boundary scans when new…
The on-demand powerful supervisor a stuck low-tier agent consults. Answers EXACTLY ONE escalated sub-question, seeing only the draft + question + a compact critique instruction — NEVER the raw file/project context (that withholding keeps the call cheap). Returns a guild.handoff.v2 envelope; the cheap/mid agent…
The mid-tier installed machinery agent that ASSEMBLES bounded context and emits run-scoped capability evidence — the executor for Learn's semantic operations and for per-lane context bundles, so a project-local role is never also the thing that assembles its own context. Writes ONLY under .guild/context/…
The generic mid-tier task-lane implementer: takes a settled plan lane and writes/refactors its code (draft, reason, build), escalating to the advisor when above its tier. The default ephemeral worker for implementation lanes that are NOT domain-recognizable. TRIGGER for "implement this lane", "build this task", "write…