agentleFS
mikeparcewskimikeparcewski/wicked-garden/skills/core/SKILL.md

wicked-garden-core

Skill9 starsChanged 7 months agoMITView on GitHub

Core utility surface for the wicked-garden plugin: help/overview, first-run setup + codebase onboarding, the wicked-* peer installer, selective state reset, the where-am-i path manifest, and structured issue filing.

  • Installs packages

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 1 is distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    If none of these is available, or Python 3 is missing, skip the script-backed step, say so, and follow the manual alternative where one is …

    line 31
  2. Very specific:

    Every prompt is classified into one or more archetypes by the UserPromptSubmit hook.

    line 131
  3. Very specific:

    Read-only query that prints a single compact manifest of the storage roots used by wicked-garden dispatches.

    line 151

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Wicked Garden Core

One entry point for the plugin's utility verbs. Run this skill inline, never as a hand-off — the setup / install / reset / report-issue actions are interactive wizards that converse with the user: your harness's native question prompt where it has one, else a numbered plain-text list and STOP-and-wait (the PLAIN_TEXT mode every ref documents). Everything runs in the parent context.

Runtime

Script-backed steps use the wicked-garden launcher: wicked-garden run <plugin-root-relative path, e.g. scripts/…> [args]. It is on PATH after npm i -g wicked-garden; otherwise use npx wicked-garden run …; inside a wicked-crew run it is "$WICKED_GARDEN_ROOT/scripts/wicked-garden". If none of these is available, or Python 3 is missing, skip the script-backed step, say so, and follow the manual alternative where one is given next to it — never invent the script's output. Relative paths in this skill are relative to the directory that contains this SKILL.md. Runtime resolution in depth — the Python/Node detection ladders, usage patterns and troubleshooting — is refs/runtime.md.

Action router

ActionTrigger phrases / argsHow
helpno args, "help", "overview"Default — show the overview below
setup"setup", "--reconfigure", "onboard this codebase"read refs/setup.md (relative to this skill's base directory) and follow it
install"install", "add peers", "install layers"read refs/install.md and follow it
reset"reset", "clear state", --all --only --keep --force --list-projects --all-projectsread refs/reset.md and follow it
where-am-i"where am I", "path manifest", --fence --envInline section below
report-issue"report issue", "file a bug", bug ux-friction unmet-outcome --list-unfiledread refs/report-issue.md and follow it

Action: help (default)

Show this overview when invoked with no action or asked for help.

The Wicked Garden Marketplace — gap-filling capabilities for modern coding-agent harnesses. It reads each prompt as one or more work-shape archetypes (v11) and applies the right rigor/gate, re-derives "done" from evidence, surfaces relationships grep can't see, and otherwise stays out of the harness's way.

Headline skills

SkillWhat it does
wicked-garden-proveRe-derive "done" from recorded evidence (the produces-gate). Its compile action emits a self-contained, vault-backed gate into any repo.
wicked-garden-coreThis skill — setup, install, reset, where-am-i, report-issue, help.
wicked-garden-deliberateCritically analyze a request before doing the work — challenge assumptions, find root causes, propose better approaches.
wicked-garden-archetypeThe v11 work-shape playbooks (below).
wicked-garden-governed-workerThe discipline every governed unit follows, by role (creator / evaluator / neutral): the repo's checks run in-tree with exit codes pasted, evaluators output-only, honest counts, no questions into a headless run.

Archetypes (v11 work-shape model)

Invoke the wicked-garden-archetype skill with a work-shape name to run its playbook. Each prompt classifies into one or more of these; run them in dependency order.

ArchetypeShape
triageclassify → routing decision
exploreframe → diverge → converge
specifyelicit → structure → validate (SMART acceptance criteria)
decidebrief → options → score → record (ADR)
shipcanary → ramp → full → soak (rollout verdict)
reviewscope → assess → findings → remediate-or-accept (hard verdict)
incidenttriage → investigate → mitigate → resolve → followup
buildplan → implement → test → review
migrateplan → expand → backfill → cutover → contract

Domain skills

Each domain is one consolidated skill that routes to its actions.

SkillDescriptionKey actions
wicked-garden-agenticDesign, review, and audit agentic AI systemsreview · design · audit · frameworks
wicked-garden-appsChange, review or accept an Apps app (a declarative UI over live data in studio): a turn's builder edits apps.json and queries/ only, from the catalog, and never calls a write tool or an action (a live read goes only through crew's broker); a distinct seat reviews the tree; a publish's QE seat judges the publish check's evidencebuild · review · qe
wicked-garden-build-guideDiscover how a repo builds and tests and write its build/test guide (.wicked/build-guide.json) for the wicked-core floor: CI config first, structured results first, a mutation probe, validated by wicked-core floor probe, proposed as a PR. Runs in setup onboardingdiscover · refresh · check
wicked-garden-dataData analysis, pipelines, ML, and ontology recommendationsanalyze · pipeline · ml · ontology
wicked-garden-demoPlan, record and review a polished product demo video — presenter script, per-chapter recordings, and contact-sheet QAplan · record · review
wicked-garden-engineeringArchitecture, code review, debugging, docs, planning, and deterministic multi-file code transformationsreview · debug · arch · plan · apply
wicked-garden-jamMulti-model brainstorming + structured council (independent second opinion)council · brainstorm · quick · revisit
wicked-garden-domainExtract a codebase's domain model — business rules + provenance, entities, requirements on the estate graph — a foundational substrate build/migrate/review/modernize all consume (none own)extractor · modeler · coverage
wicked-garden-draftThe quality floor for document deliverables drafted from a repo (brochures, one-pagers, pages, decks): contrast ≥ 4.5:1 and print text ≥ 7pt, the RENDERED page count honours the brief, every number/URL cited, mocks labelled, no placeholders — with one self-check run before "done"self-check · contrast · pages · claims
wicked-garden-editor-scaffoldBuild an artifact editor plugin for wicked-studio: scaffold a spec-2 pack with one self-contained wicked.editor/1 editor, ask only for the permissions it uses, re-pin its hash, run the conformance cases in headless Chromium against studio's plugin host, then install it into a wicked-crew daemon's editors (the daemon re-runs the conformance)scaffold · pin · conformance · install
wicked-garden-learnLearn a topic or an event from outside sources into the repository as KNOWLEDGE: research with cited sources, a claims ledger (every claim with its URL and observed_at), a chat walkthrough, text-only skills that cite the claims; a refresh proposes only changes (confirm, supersede, contradict, retire), never a silent delete; index merged knowledge into estate. Report mode (sourced-report: research, competitive, due diligence, analyze, critique, estimate) reports from documents, picked sources and the web, then a fidelity check re-checks every claim; it changes nothingscope · research · synthesize · walkthrough · author · review · fidelity · check · index
wicked-garden-mcpUse an MCP tool from a governed wicked-crew unit on any seat: list the tools your unit may try, then call one through the broker (policy-judged, recorded, secrets never handed out)list · call
wicked-garden-mcp-scaffoldBuild an MCP server the wicked-crew broker can govern: scaffold a TypeScript server (fastmcp, OpenTelemetry, loglayer — the default) or a zero-dependency one-file stdio server (Node or Python), generate its tools from an OpenAPI document through wicked-crew's conversion service, annotate them honestly, probe initialize + tools/list with each tool's broker class, then install it for running (registry + CLI configs, idempotent by key)scaffold · openapi · probe · install
wicked-garden-memCross-session memory + knowledge over wicked-estate: store/recall learnings, cited answers, document ingest (binary via vision), session capturestore · recall · answer · review · forget · maintain · ingest · capture
wicked-garden-personaDefine and invoke named personas to perform work with a specific lensas · define · list
wicked-garden-platformSecurity, infrastructure, compliance, CI/CD, incidents, traces, and plugin diagnosticssecurity · compliance · incident · health
wicked-garden-productRequirements, customer feedback, strategy, UX, accessibility, and design reviewelicit · acceptance · analyze · strategy · ux-review
wicked-garden-qeEvidence-gated quality engineering: strategy, scenario authoring, execution, independent verdicts, ledger insight, and the 3-agent acceptance pipelinesetup · plan · author · execute · review · insight · accept
wicked-garden-repo-learnLearn an unfamiliar repo (churn → hotspots → read the load-bearing intersection) and capture BOTH memories and policies as inert estate proposals a human reviewschurn · hotspots · read · capture
wicked-garden-searchStructural code search, lineage, blast-radius, and codebase intelligenceblast-radius · lineage · hotspots · service-map · index

Cross-session memory + knowledge is the wicked-garden-mem domain (wicked-estate is the engine): store / recall / answer / ingest replace the retired wicked-brain skill cluster. Relationship graphs <!-- historical --> (blast-radius / lineage / hotspots) live in the wicked-estate MCP (BlastRadius / Lineage / TraverseGraph / RankHotspots, ADR 0005) — the wicked-garden-search skill wraps them; symbol lookup is the estate MCP SearchEntity tool.

Quick start

  • Orient: invoke this skill's where-am-i action.
  • Build a feature: wicked-garden-archetype with build "add a user authentication system".
  • Re-derive "done" from evidence: wicked-garden-prove.
  • Review code: wicked-garden-engineering with review ./src.
  • Independent multi-model second opinion: wicked-garden-jam with council "should we adopt event sourcing here?".
  • Search code relationships: estate MCP SearchEntity for the symbol, then wicked-garden-search with blast-radius src/payments.py.
  • Store a decision: wicked-garden-mem (store action).

How it works

  1. Every prompt is classified into one or more archetypes by the UserPromptSubmit hook; each archetype owns its own phase shape, HITL discipline, and cost band (steering, not a fixed pipeline).
  2. Specialist domains (engineering, platform, product, qe, data, agentic, jam, search) provide deep expertise the harness routes into. Context is assembled on demand (pull-model) via the wicked-garden-mem skill — there is no per-prompt push.
  3. wicked-garden-prove re-derives an archetype's "done" through the evidence gate rather than trusting a "tests pass" claim.
  4. State persists across sessions via wicked-estate memory, search indexes, the event log, and native tasks.

Getting more help

Invoke any domain skill with an action name (e.g. wicked-garden-engineering with review). Invoke wicked-garden-core with no action to return to this overview.

Action: where-am-i

Read-only query that prints a single compact manifest of the storage roots used by wicked-garden dispatches. Subagents should include "invoke wicked-garden-core where-am-i first" as a directive instead of hand-enumerating paths — it costs fewer tokens and closes a class of path-mismatch bugs. Provenance: Issue #576.

Arguments: --json (JSON manifest, default) · --fence (wrap the JSON in a fenced json block for paste) · --env (substitute env-var forms — the plugin-root variable — where the corresponding environment variable is present).

Invoke the helper script and stream its stdout to the user verbatim. This action is a thin wrapper — do not re-interpret the manifest.

wicked-garden run scripts/where_am_i.py "$@"

Output shape:

{
  "plugin_root": "/abs/path",
  "source_cwd": "/abs/path",
  "active_project_id": "string-or-null",
  "project_artifacts": "/abs/path/to/projects-or-specific-project",
  "bus_db": "/abs/path"
}

Any field that cannot be resolved emits null and logs a one-line note to stderr. The script never raises and is safe to invoke from any cwd. Graceful degradation: a missing plugin-root variable falls back to the checkout inferred from the script location; missing bus DB emits "bus_db": null; no active crew project emits "active_project_id": null and points project_artifacts at the crew projects domain root.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.