Help us improve
Share bugs, ideas, or general feedback.
Use when working as Sonzai internal staff alongside sonzai-sdk or full-auto. Layers awareness of the Sonzai monolith (sonzai-ai-monolith-ts — contextengine, platform/api, ai-character-service, deploy) and SDK source workspace ($SONZAI_WORKSPACE/sonzai-sdk/) onto the public skills so wizard recommendations and autonomous builds can be cross-checked against server source instead of just the OpenAPI snapshot, and QA cycles can trace SDK calls into server handlers.
npx claudepluginhub sonz-ai/sonzai-claude-skill --plugin sonzai-internal-staffHow this skill is triggered — by the user, by Claude, or both
Slash command
/sonzai-internal-staff:sonzai-internal-staffThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Internal-only companion to `sonzai-sdk` and `full-auto`. Layers awareness of the monolith repo (`sonzai-ai-monolith-ts`) and the broader Sonzai workspace so internal staff can dogfood the public skills with full server-side context.
Creates p5.js generative art with seeded randomness, noise fields, and interactive parameter exploration. Use for algorithmic art, flow fields, or particle systems.
Share bugs, ideas, or general feedback.
Internal-only companion to sonzai-sdk and full-auto. Layers awareness of the monolith repo (sonzai-ai-monolith-ts) and the broader Sonzai workspace so internal staff can dogfood the public skills with full server-side context.
This plugin is install-time gated. The marketplace lists it as category: internal with a description that flags it as useless without monolith access. If you somehow have this skill loaded but are NOT Sonzai staff with read access to the private monolith repo, uninstall: /plugin uninstall sonzai-internal-staff@sonz-ai.
The public skills (sonzai-sdk, full-auto) MUST remain tenant-agnostic and free of platform-internals references. This skill is the safe place to put internal pointers.
This skill exists so Sonzai internal staff can:
full-auto against dev/staging/production with real API keys, real workspace context, and the ability to trace SDK calls into server handlers when QA cycles failsonzai-sdk (wizard) while cross-checking recommendations against the live SDK source at $SONZAI_WORKSPACE/sonzai-sdk/ AND the server source at $SONZAI_WORKSPACE/sonzai-ai-monolith-ts/services/contextengine/ instead of just the OpenAPI snapshotStaff workspace layout (where you read from — exact path varies per machine):
$SONZAI_WORKSPACE/ # set this env var; never hardcode a path
├── sonzai-claude-skill/ # this skill repo
├── sonzai-sdk/ # SDK source (sonzai-python, sonzai-typescript, sonzai-go, sonzai-openclaw, ...)
├── sonzai-ai-monolith-ts/ # monolith (contextengine, platform/api, ai-character-service, deploy)
├── sonzai-landing/ # public docs source (sonz.ai/docs)
└── ... # tenant projects — out of scope
Resolving $SONZAI_WORKSPACE (in priority order):
$SONZAI_WORKSPACE if set and existssonzai-sdk/ and sonzai-ai-monolith-ts/$HOME/code/sonzai/, $HOME/work/sonzai/, $HOME/dev/sonzai/, $HOME/src/sonzai/SONZAI_WORKSPACE to the directory containing sonzai-sdk/ and sonzai-ai-monolith-ts/." and exitRead all three of sonzai-sdk/, sonzai-ai-monolith-ts/, and the public skill (sonzai-sdk / full-auto under sonzai-claude-skill/plugins/sonzai-sdk/skills/) when doing internal work. See workspace-pointers.md for the full layout.
full-auto Phase 0 (drift check): in addition to the live https://api.sonz.ai/docs/openapi.json, ALSO read:
$SONZAI_WORKSPACE/sonzai-ai-monolith-ts/services/platform/api/docs/ for the most current schema (live may be one deploy behind)$SONZAI_WORKSPACE/sonzai-sdk/sonzai-{python,typescript,go}/ to confirm the SDK actually exposes the symbolfull-auto Phase 5 (QA loop): when a fix cycle fails, optionally tail server logs from the local monolith to diagnose. Stop short of editing platform code — that's a separate task.sonzai-sdk wizard: when wizard derivation hits a capability question, verify against:
$SONZAI_WORKSPACE/sonzai-ai-monolith-ts/services/contextengine/domain/entity/agent.go (server truth)$SONZAI_WORKSPACE/sonzai-sdk/sonzai-{python,typescript,go}/ (caller truth)plugins/sonzai-sdk/skills/; don't drag internal pointers into themdeploy/ scripts directly; this skill doesn't gate deploymentworkspace-pointers.md — where things live in the broader Sonzai workspace: SDK source (sonzai-sdk/), monolith (sonzai-ai-monolith-ts/), public docs (sonzai-landing/). Paths, key files, source-of-truth ordering.sdk-server-side-debugging.md, staging-workflow.md, dogfooding-full-auto.md — add when friction demands)If you also have sonzai-sdk or full-auto loaded (the normal case for internal dogfooding):
.env files and secrets are out of bounds for this skill's output.plugins/sonzai-internal-staff/skills/sonzai-internal-staff/ only.git push from this skill's context to the monolith repo. (full-auto already enforces this for target repos; this rule extends it to the monolith.)This plugin ships in lockstep with sonzai-sdk from the same repo (github.com/sonz-ai/sonzai-claude-skill). See the repo's CHANGELOG.md for release history.