fastloom-reference

fastloom reference docs

This skill bundles the full fastloom documentation. When the user asks about a fastloom concept, read the relevant file from docs/ next to this SKILL.md and ground your answer in it.

Doc map — which file answers which question

Topic	File
Bootstrapping a new service, the app.py / settings.py contract, the launch CLI	docs/quickstart.md
App declarative model, startup order, lifespan composition, LauncherSettings	docs/launcher.md
Composing Settings, Configs / TC.general, tenants.yaml structure, env-var fallbacks	docs/settings.md
TC[tenant] / TC.set, cache→Mongo→YAML resolution, tenant dependency sources, TenantMixin	docs/tenant.md
JWTAuth / UserClaims, IAM topology (Casdoor/Keycloak + sidecar), /introspect and /acl	docs/auth.md
Beanie models, mixins (CreatedUpdatedAtSchema), BasePaginationQuery, BaseDocumentSignal	docs/db.md
RabbitSubscriber, publishers, subscribers, retry/backoff DLX topology, init_streams	docs/signals.md
RedisHandler, BaseCache / BaseTenantSettingCache, HostTenantMapping, RedisGuardGate	docs/cache.md
init_healthcheck, auto-registered checks, custom handlers	docs/healthcheck.md
InitMonitoring, Instruments enum, infer_instruments, OTel sampling, Sentry/Logfire	docs/observability.md
CustomI18NException, i18n_exception_handler, get_template, lang_dict	docs/i18n.md
FileIn / FileObject / FileMessage / FileField, outbox-matching pattern	docs/file.md
MCPSettings, FastMCP mount, bearer-forwarding auth	docs/mcp.md
init_app / TC / auth fixtures, assert_deep_diff, testcontainers create_container	docs/test.md
Cross-cutting conventions: typing rules, SelfSustaining, optional-import idiom, validators, naming, dedupe	docs/conventions.md

How to use this skill

When the user asks a question that touches one of the topics above:

1. Read the matching file with the Read tool. Paths are relative to this skill — e.g. Read("docs/auth.md").
2. Answer from the doc, citing docs/<file>.md in your reply so the user can verify.
3. If the question spans multiple files (e.g. "how does auth flow into per-tenant settings?"), read each relevant file and synthesize.
4. If the user is editing a fastloom-based service and asks something underspecified ("how do I add a route?"), point them at the right doc and offer to invoke the dedicated scaffolding skill (e.g. /fastloom-sdk:add-fastloom-route).

What's bundled

The docs/ directory next to this SKILL.md is a symlink to the canonical docs/ of the fastloom repository. At install time, Claude Code dereferences the symlink and copies the actual files into the plugin cache — the user gets a snapshot of the docs current to the installed plugin version.

To get newer docs, the user runs /plugin update fastloom-sdk@fastloom (when a newer version is published).

Important: don't substitute training-data knowledge for what's in docs/

Fastloom moves. Field names, capability mixins, and conventions shift between versions. If you're tempted to answer a fastloom question from memory, stop and read the doc first. The doc is the source of truth — your training data is not.

Cross-referencing the scaffolding skills

The other four fastloom-sdk skills (scaffold-fastloom-service, add-fastloom-route, add-rabbit-subscriber, audit-fastloom-settings) embed key rules inline, but when a user asks "why does this scaffold do X", come back here and read the matching doc.