Every Claude Code skill that exists anywhere in FLME — what it is for, which repo owns it, and whether it is actually live. A skill is a packaged procedure an agent session can invoke by name; .claude/skills/ is the directory Claude Code discovers them from.
This page is an index, not a summary. Each description below is that skill's own first sentence, lifted verbatim from its SKILL.md frontmatter and reformatted, never rewritten. For what a skill actually does, open the skill — a paraphrase here would drift out of sync with the real thing and nobody would notice.
Counts. 60 skills live in the six repos that have an owning agent, listed individually below. A further 38 live in client & deployment repos that have no owning agent — counted at the bottom of this page, deliberately not listed row by row. Total across FLME: 98.
These live in flme-agents/.claude/skills/ and are symlinked into ~/.claude/skills/ by scripts/install-skills.sh, which is what makes them invokable from any repo rather than only from flme-agents. Owned by Nora. They are process skills: how work is handed off, shipped, posted and reported, rather than anything about one product.
| Skill | Its own first line | Status |
|---|---|---|
daily-digest | Use when asked for a daily digest, status pull-up, agent status report, "how's everyone doing," or a standup-style summary across all FLME agents ("give me the daily pull-up," "what's everyone been up to," "status report on the agents"). | Live |
decision-brief | Use when Jake (or any human lead) asks where the work stands and what is waiting on them — "where are we," "what do I need to decide," "status and decisions," "what's blocked on me," "give me the rundown," "what's waiting for my call." Produces a two-part brief: status items that need no action, then decisions that belong to the human, each with options, a recommendation, and the cost of not deciding. | Live |
delegating-work | Use when deciding whether to hand work to a separate worker session, whether a brief needs a blind readability check before handoff, or whether to escalate to a higher-capability advisor. | Live |
playwright-browser-automation | Use when the user asks to verify, check, or test a website/page/flow using Playwright ("verify this with playwright", "check this page works", "automate this read-only check"), or to drive an authenticated admin dashboard or deployment-config panel using Playwright ("configure Cloudflare using playwright", "drive this dashboard with playwright", "automate this logged-in flow"). | Live |
project-worktrees | Run a larger project across several agents and repos using one git worktree per agent, all on a branch named after the project — so each agent commits and pushes freely without coordinating, and the project lands on main as one periodic merge instead of a stream of approvals. | Live |
refine-prompt | Use when asked to refine, tighten, polish, or clean up a prompt before it gets sent to an LLM or subagent ("refine this prompt," "polish this for a subagent," "tighten the wording on this," "check this brief is clear enough before I hand it off," "is this prompt clear enough?"). | Live |
shipping-changes-safely | Take a change from "written" to "live" without breaking what is already there — knowing what a push actually does in this repo, isolating the work, verifying it with a gate that can actually see the change, landing it with a merge proof, confirming the live result, and having a rollback target ready before you need it. | Live |
slack-post | Use when an FLME agent needs to post to Slack as itself — a Cadence project update, drain summary, or thread reply; the daily digest; any agent-authored message in a `#_bot_` channel ("post this to #_bot_cadence," "reply to John in Slack," "send the digest"). | Live |
splitting-an-oversized-file | Use when a file has grown too large to be read in full and its history or bulk must be moved out — a status file near a size ceiling, a log nobody can open, any document whose own instructions say to read it whole and which no longer can be. | Live |
ux-security-review | Use when reviewing a change to `ux-template` (the FLME UI generator) or a UI it generated, for security judgment calls the CI scanners cannot make: "security review this diff," "review this PR for security," "check this generated instance," "is this auth change safe," or after a real build that touched the generator, `src/auth/`, `src/lib/`, the journey layer, `src/proxy.ts`, deploy packaging, the OpenAPI snapshot, or any `NEXT_PUBLIC_*` value. | Live |
These are discoverable only inside their own repo — Claude Code loads project-level skills from the session's own repo, so a ux-template skill is invisible to a session working in flme-release-coordination. Three of them (design-ux-from-reqdoc-drafts, build-ux-from-reqdoc-drafts, package-ux-deployment) also appear on the Engineering Reference as pipeline stages; that page describes the stage, this one indexes the skill.
ux-template — UI pipeline + design taste 21 skills owner: Wren| Skill | Its own first line | Status |
|---|---|---|
bind-reqdoc-drafts-live-api | Points one reqdoc-drafts client build at a real live backend instead of synthetic mock data: fetches a fresh OpenAPI schema snapshot, resolves every noun's own field bindings against it (bound/unbound/shape-mismatch, per noun and per field — the real Track A/B/C/D logic in resolveLiveApiBinding.ts), and writes the resolved dataMode/apiBaseUrl/ keycloakIssuer/keycloakClientId/apiSchemaSnapshotPath fields into that client's own ux-designs/<CLIENTTOKEN>/<token>.ux-design.json. | Live |
bootstrap-deployment | Configures a new UI deployment's own identity — UI_PROJECT_NAME and everything it cascades into (Docker container name, network name, dev port) — away from ux-template's own placeholder defaults, and confirms the paired-backend identity recorded by derive-deployment's .backend-deployment marker resolved correctly. | Live |
brandkit | Premium brand-kit image generation skill for creating high-end brand-guidelines boards, logo systems, identity decks, and visual-world presentations. | Live |
build-ux-from-reqdoc-drafts | Instantiates real React/TypeScript UI — a noun's real master-detail Standard Presentation page, and role-scoped Dashboards reproducing the real GearDesk wireframe's own shell directly (topbar, notification/ profile popovers, tab bar, item-card queues, browse tables) — all with a live entity-mapping hover tooltip and a runtime look-and-feel switcher (eight real presets, swappable with no regenerate) — from a CONFIRMED design decision for a reqdoc-drafts client package (a richer, more detailed markup source than the plain live-API ReqDoc pipeline generate-client-ui consumes; e.g. GearDesk under /Users/jakeyara/FLME/dktest/reqdoc-drafts). | Live |
derive-deployment | Creates a new client deployment repo from ux-template, pinned to a specific tagged release with shared git history (not a disconnected GitHub "template repository" copy), wires up an `upstream` remote pointing back at ux-template so a future `pull-template-release` can merge in newer releases, records the pin in a `.template-release` marker file, optionally records which ReqDoc client this deployment is for in a `.reqdoc-client` file (so `generate-client-ui` can default to it instead of needing the client named every run), records which backend deployment this UI talks to in a `.backend-deployment` file, and then hands off to `bootstrap-deployment` to configure the new repo's own identity (Docker project name, dev port, etc), finishing with a pointer to `flme-release-coordination` for what to do next. | Live |
design-taste-frontend | Anti-slop frontend skill for landing pages, portfolios, and redesigns. | Live |
design-taste-frontend-v1 | The original v1 taste-skill, preserved for projects depending on its exact behavior. | Live |
design-ux-from-reqdoc-drafts | Guides a human through DESIGNING a reqdoc-drafts client's own UX shape — whether the screen set actually delivers the package's own stated purpose end-to-end, screen role (standalone/embedded/cross-reference- only/excluded), home/entry point, index composition, look-and-feel default — before any code generates, producing a reviewable, resumable ux-designs/<CLIENTTOKEN>/<token>.ux-design.json artifact with live visual feedback throughout (a real /studio/draft/<token> preview a human can click through, never abstract JSON review). | Live |
full-output-enforcement | Overrides default LLM truncation behavior. | Live |
generate-client-ui | Wipes and regenerates a client's UI from markup, verified against a real live schema, with zero manual steps. | Live |
gpt-taste | Elite UX/UI & Advanced GSAP Motion Engineer. | Live |
high-end-visual-design | Teaches the AI to design like a high-end agency. | Live |
image-to-code | Elite website image-to-code skill for Codex. | Live |
imagegen-frontend-mobile | Elite mobile app image-generation skill for creating premium, app-native screen concepts and flows. | Live |
imagegen-frontend-web | Elite frontend image-direction skill for generating premium, conversion-aware website design references. | Live |
industrial-brutalist-ui | Raw mechanical interfaces fusing Swiss typographic print with military terminal aesthetics. | Live |
integrate-client-ui | Cross-checks a client's UI Prototype markup against a live OpenAPI schema and produces a readable report of what will and won't generate cleanly, before generate-client-ui wipes and regenerates src/components/generated/. | Live |
minimalist-ui | Clean editorial-style interfaces. Warm monochrome palette, typographic contrast, flat bento grids, muted pastels. | Live |
package-ux-deployment | Packages a client's CONFIRMED UX design as a real, self-contained Next.js Docker service inside a paired backend deployment repo, and attaches it to that repo's own docker-compose.yml on the next free slot of its reserved UX port band — verifying the design artifact isn't stale, scaffolding the standalone package, emitting the client's real screens rooted at "/", and patching the compose file with a service block whose every `${...}` expression is copied verbatim out of that file's own existing services (so the new frontend cannot drift onto a different backend or realm than the one already there). | Live |
redesign-existing-projects | Upgrades existing websites and apps to premium quality. | Live |
stitch-design-taste | Semantic Design System Skill for Google Stitch. | Live |
backend-template — Backend codegen + template releases 13 skills owner: Doug / backend-template| Skill | Its own first line | Status |
|---|---|---|
abt-binding-to-engine | MOVED — this skill now lives in the FLM-Eng/flme-framework repo, not here. | Moved away |
audit-upstream-fix | No description — this skill has no YAML frontmatter, so Claude Code cannot discover it. | No frontmatter |
bootstrap-deployment | Configures a new deployment's identity — the Postgres database name, the Keycloak realm and its 3 clients, every Docker container/network/volume name, the Temporal task queue, and the product name shown in the web UI — away from "flme" (the original reference deployment's name). | Live |
build-client-api | MOVED — this skill now lives in the FLM-Eng/flme-framework repo, not here. | Moved away |
catalog-to-framework | Clears every entity/workflow that noun-to-entity, verb-to-workflow, and their supporting scripts (scaffold-entity.js, generate-json-path-indexes.js, generate-types.js, generate-index.js, generate-activities.js) have produced out of this backend, leaving only the hand-maintained framework underneath — entity-kit, the generic repositories, EntityCrudController, the entity-form UI engine, and so on. | Live |
cut-template-release | Tags a new semver release (vMAJOR.MINOR.PATCH) of this backend-template repo itself, and — for a bug fix that needs to reach an older major line still running in derived deployment repos — lazily creates or reuses a `release/vN` branch to carry an isolated backport, so a deployment pinned to that line can pull just the fix without dragging in newer, unrelated major-line changes. | Live |
derive-deployment | Creates a new client deployment repo from backend-template, pinned to a specific tagged release with shared git history (not a disconnected GitHub "template repository" copy), wires up an `upstream` remote pointing back at backend-template so `pull-template-release` can later merge in newer releases, records the pin in a `.template-release` marker file, optionally records which ReqDoc client this deployment is for in a `.reqdoc-client` file (so `generate-client-api` can default to it instead of needing the client named every run), and then hands off to `bootstrap-deployment` to configure the new repo's own identity (Postgres db name, Keycloak realm, Docker names, product name, etc), finishing with a pointer to `flme-release-coordination` for what to do next. | Live |
noun-to-entity | MOVED — this skill now lives in the FLM-Eng/flme-framework repo, not here. | Moved away |
provision-client-deployment | The idempotent, whole-pipeline wrapper for standing up (or advancing) one client's real deployment: derive-deployment (fork the repo) → bootstrap-deployment (identity config) → generate-client-api (scaffold the API from the frozen flme-release-coordination release). | Live |
pull-template-release | Pulls a newer backend-template release into an already-derived deployment repo: fetches tags from the `upstream` remote `derive-deployment` set up, shows what's newer than the repo's current `.template-release` pin, lets the user pick a specific target tag (a same-major tag is a routine pull; a higher-major tag is a deliberate upgrade, per cut-template-release's semver convention), merges it on a dedicated branch, and updates the pin. | Live |
review-spec | MOVED — this skill now lives in the shared flme-release-coordination repo, not here. | Moved away |
upstream-fix | Carries a bug fix discovered and made in a derived deployment repo back into backend-template itself, using the shared git history `derive-deployment` set up: fetches the fix commit straight from the deployment repo's URL (no persistent remote needed), splits it out if it's tangled with client-specific changes, cherry-picks the isolated fix onto a backend-template branch, and opens a normal PR. | Live |
verb-to-workflow | MOVED — this skill now lives in the FLM-Eng/flme-framework repo, not here. | Moved away |
flme-release-coordination — The release process 12 skills owner: Sam| Skill | Its own first line | Status |
|---|---|---|
abandon-release | Closes out a client's in-progress release that's stuck for a real reason (scope changed, not just "needs another fix round") — step 6 (Freeze)'s own sibling action, not a numbered step itself. | Live |
browse-consolidated-findings | Reads an already-rendered findings-report.json (produced by consolidate-spec-findings, shaped to the FLME Requirements Issue Exchange Standard) and lets the caller explore it — print the full report, group by severity/response-lane/entity/rule, filter to just blocking or just auto-fixable (LLM-lane) findings, or show one finding's full detail. | Live |
browse-validation-report | Reads an already-generated ui_requirements_validation.py report (the UX requirements validator's own raw markdown output — Executive Summary, Integrity Check Scorecard, Breakdown by Document Category, Errors, Warnings) and lets the caller explore it — full report, by check, by category, errors only, warnings only, a single file's full detail, or a keyword search across finding text. | Live |
build-client-closure | Resolves a ReqDoc client application package's full Core→Domain→Client dependency closure — every noun, verb, and transaction it actually needs, classified and dependency-ordered — from its own Activity Inventory. | Live |
consolidate-spec-findings | Parses a release iteration's gate/integration reports (review-spec.md-style prose reports, cross-check-bindings.md-style table/issue-code reports, or any other registered report format), cross-links same-root-cause findings across sources without hiding source-specific detail, classifies each as auto-fixable-proposable or needs-human, and renders one consumable findings report shaped to the FLME Requirements Issue Exchange Standard's common envelope (ReqDoc, Specstrum/Process/Standards/) — noting client, markup source, accepted baseline, and generated API version per finding. | Live |
freeze-release | Step 6 (Freeze) of this repo's own release process (top-level README) — checks a client's in-progress release's own latest iteration for freeze-readiness (zero Blocking findings across every gate that should have run, plus every Flagged finding surfaced for visibility even though it doesn't block), confirms with the caller, then writes a permanent frozen marker in the release's own README.md. | Live |
render-release-swimlane | Renders a release's own multi-iteration history as a swimlane — one arrow-chain rail per iteration, newest first, all 10 steps, a published glyph legend — and this *is* the rail now, not a separate section beside the old single flat one. | Live |
reopen-release | Flips an already-frozen release back to `in-progress` when step 9 (Integrate)'s own gate found a Blocking finding disposed `spec-inadequate` — the reopen trigger named as "designed, not yet built" in `integration-gate-registry.md` and canonical-backlog item 39, now real. | Live |
review-spec | Reviews a ReqDoc noun/verb/transaction spec — or a whole client package's resolved dependency closure — for gaps, ambiguity, or missing information BEFORE generation, using a checklist of what noun-to-entity/verb-to-workflow/abt-binding-to-engine actually need to scaffold something correctly. | Live |
run-release-iteration | Runs steps 2 (Snapshot), 3 (Resolve), and 4 (Review) of this repo's own release process (top-level README) end to end for one client, given just its name — cuts or advances the release branch, freezes the tree-shaken closure as this iteration's manifest.json/manifest.md, then runs every shipped gate in gate-registry.md against it and reports Blocking/Flagged results. | Live |
run-retrospective | Runs step 11 (Retrospective) of this repo's own release process (top-level README) for one client release — reads every step's own accumulated findings/<step-slug>.md plus each iteration's gates/ and integration/ output, classifies each real, generalizable finding by audience, and writes one synthesized report per team into clients/<client>/<release>/retrospective/<team>.md. | Live |
snapshot-release | Cuts a new versioned ReqDoc release branch for a client, or advances an already-in-progress release to its next review iteration — step 2 (Snapshot) of this repo's own release process (top-level README). | Live |
flme-deployments — Deployment provisioning 2 skills owner: Skipper| Skill | Its own first line | Status |
|---|---|---|
provision-tunneled-deployment | Sets up the git-push-triggered, Cloudflare-Tunnel-served deployment mode for a client under flme-deployments: detects whether the client is mock-data or live-API (a live-API client needs its api and keycloak services publicly reachable too, not just its UX, because their browser-facing URLs are baked into the UX's build), generates a cloudflared ingress config mapping <client>.flme.ai (and, for a live-API client, <client>-api.flme.ai and <client>-keycloak.flme.ai) to the deployment's local ports, generates a self-hosted-runner GitHub Actions workflow at the repo root that runs `docker compose up -d --build` on push to main, and wires the Keycloak realm template so the deployment's own already-running keycloak-realm-reconcile mechanism picks up the new public hostname as a trusted redirect URI. | Live |
run-deployment-locally | Runs and operates any generated deployment directory under flme-deployments (identified by its five identity-carrying files — .env.example, package.json, deployment-adoption.json, README.md, CLAUDE.md — not hardcoded to pawstop) as a local Docker Compose stack on the operator's own Windows/Mac machine: brings it up, tears it down, rebuilds a service after a source or build-arg change, polls every service's own healthcheck until ready, and reports back the live host URL for each published service. | Live |
ReqDoc — Requirements vault 2 skills owner: John / ReqDoc| Skill | Its own first line | Status |
|---|---|---|
browse-consolidated-findings | HOSTED ELSEWHERE, runnable directly from right here — this skill's real content lives in the sibling flme-release-coordination repo (.claude/skills/browse-consolidated-findings/SKILL.md). | Pointer stub |
browse-validation-report | HOSTED ELSEWHERE, runnable directly from right here — this skill's real content lives in the sibling flme-release-coordination repo (.claude/skills/browse-validation-report/SKILL.md). | Pointer stub |
These repos carry their own copies of the backend pipeline's skills, and the copies have diverged from backend-template's originals rather than tracking them — bootstrap-deployment is 23,569 bytes in browser against 21,955 in backend-template, and cut-template-release is nearly double. None of these repos has an owning agent.
They are counted here rather than listed because a per-client fork's row would go stale every time that client is regenerated, and ~38 such rows would bury the 60 skills anyone actually invokes. The boundary is the point: this page would otherwise read as complete while silently omitting them.
| Repo | Skills | Relationship |
|---|---|---|
browser | 13 | Diverged per-client variants of backend-template's set. No owning agent. |
petstore | 12 | Diverged per-client variants of backend-template's set. No owning agent. |
dktest | 11 | Diverged per-client variants of backend-template's set. No owning agent. |
backend | 2 | Diverged per-client variants of backend-template's set. No owning agent. |
One repo referenced but not present. Five backend-template skills are tombstones reading “MOVED — this skill now lives in the FLM-Eng/flme-framework repo.” That repo is not checked out on this machine and does not appear on the Repo Atlas. Its skills are therefore not counted anywhere on this page.