phahadek

deploy

Deploy a managed project to production, executed strictly from that project's structured deploy playbook — confirm-gated, never improvised. Use when the user says "deploy", "deploy <project>", "ship <project> to prod", "run the deploy", or "push <project> live". Reads the project's context.md + the deploy playbook it points to (e.g. a repo README "Deploy" section), presents the full ordered plan, then runs it one step at a time, pausing before each production-mutating step.

Web & Frontend Listed

design

Run a Design Execution session for a milestone. Loads full project context deterministically (via design-load.mjs), investigates the open questions in each Ready / In-Progress 📐 Design and 📋 Planning task, proposes a thematic execution order for human sign-off, then walks each task one open-question at a time — debating trade-offs, locking decisions, updating architecture pages, and filing follow-on 🔲 Backlog Code tasks. Use when the user says "design", "design session", "let's run a design session for milestone X", "execute the design tasks", or starts a Design Execution session. Requires a grooming manifest in the central config tree (config/projects/<dir>/grooming.json), shared with the /groom skill.

gate

Run the Manual Verification Gate for a milestone against the gate-state API (readiness → pull runnable items by tier → human disposition → record). Use when the user says "run the gate", "gate check", "verify the gate for milestone X", or starts a Manual Verification Gate pass. Requires the gate-verification service's gate_item / gate_item_event tables and the backend's /api/gate/* routes to be live for the target milestone.

groom

Run a Backlog Grooming session for a milestone. Loads full project context deterministically (via the backend's GET /api/groom-context route), explores the code regions tasks touch with a git-fresh cache, presents 🔲 Backlog tasks in batches for human sign-off, and marks them 🗂️ Ready. Use when the user says "groom", "grooming session", "let's groom milestone X", "bring the backlog to Ready", or starts a Backlog Grooming session. Requires a grooming manifest in the central config tree (config/projects/<dir>/grooming.json).

ops

Execute a milestone's interactive 🔧 Operational and 🔎 Investigation tasks — the judgment-bound work that is NOT auto-dispatched. Two modes in one pass: Operational changes production/environment state through a sanctioned surface (config/catalog/entity authoring, backfills, alarm config, dependency installs, session/machine setup); Investigation produces a defensible decision from live data (diagnoses, spikes, zero/anomaly triage) and files follow-on tasks. Enforces: autonomous runs STAGE (never verdicts, never ✅ Done), read-only diagnosis before any write, falsify every "it's fine", audited read-modify-write, reconcile+capture, "Done ≠ deployed ≠ seeded ≠ working". Records durable research to a staging journal. Use when the user says "work the ops tasks", "run ops/tooling for milestone X", "work the operational/investigation tasks", or starts an overnight or interactive ops run on a project's task board.

Code & Development Listed

sync-guidelines

Re-vendor EVERYTHING this repo deploys into a live tree — the universal guideline docs (task-writing.md, procedures.md), the /groom /design /ops /deploy /wrap /sync-guidelines /gate skill directories, the sanctioned route-client scripts, and the config-tree hook — via a Claude-led three-way merge, never a file overwrite. The repo copy is the upstream source; the live copy may carry local edits (a vendored skill refined in place, an emergency script hotfix) that must be preserved. Use when the user says "sync guidelines", "re-vendor the skills", "deploy the grooming skills", "integrate the guideline updates", "pull the procedures/task-writing changes into config", or after merging a PR that changed config-template/, skills/**, scripts/*.mjs, or packages/backend/scripts/*.mjs.

wrap

Wrap up a session: scan it for durable information that was produced or decided but never persisted to its proper home, surface each item with a proposed home, and — on your confirmation — write it there. Use when the user says "wrap", "wrap up", "wrap the session", "let's close out", or before ending a long working session. Nothing important should die with the session.