wicked-garden-mcp
Skill9 starsChanged 7 months agoMITView on GitHub
Use an MCP tool from a governed wicked-crew unit, on any seat: list the tools your unit may try, then call one through the wicked-crew broker, which holds the server connections and secrets, judges every call by steering policy and records it. One command on every CLI; no MCP configuration.
- Installs packages
- Sends data out
In this repository142 other files it gives its agents.
- AGENTS.mdAGENTS.md
- .claude/CLAUDE.mdCLAUDE.md
- plugin-json-schemaSkill
- readme-style-guideSkill
- releaseSkill
- scaffoldSkill
- wicked-garden-agentic-agentic-patternsSkill
- wicked-garden-agentic-architectSkill
- wicked-garden-agentic-context-engineeringSkill
- wicked-garden-agentic-frameworksSkill
- wicked-garden-agentic-performance-analystSkill
- wicked-garden-agentic-review-methodologySkill
- wicked-garden-agentic-safety-reviewerSkill
- wicked-garden-agenticSkill
- wicked-garden-agentic-trust-and-safetySkill
- wicked-garden-appsSkill
- wicked-garden-archetypeSkill
- wicked-garden-build-guideSkill
- wicked-garden-classifySkill
- wicked-garden-coreSkill
- wicked-garden-data-engineerSkill
- wicked-garden-dataSkill
- wicked-garden-deliberateSkill
- wicked-garden-demoSkill
- wicked-garden-domain-coverageSkill
- wicked-garden-domain-extractorSkill
- wicked-garden-domain-modelerSkill
- wicked-garden-domainSkill
- wicked-garden-draftSkill
- wicked-garden-editor-scaffoldSkill
- wicked-garden-engineering-api-documentarianSkill
- wicked-garden-engineering-architectureSkill
- wicked-garden-engineering-backendSkill
- wicked-garden-engineering-debuggingSkill
- wicked-garden-engineering-docsSkill
- wicked-garden-engineering-frontendSkill
- wicked-garden-engineering-integrationSkill
- wicked-garden-engineering-large-scale-migrationSkill
- wicked-garden-engineering-migration-engineerSkill
- wicked-garden-engineering-patchSkill
- wicked-garden-engineeringSkill
- wicked-garden-engineering-solution-architectSkill
- wicked-garden-engineering-system-designSkill
- wicked-garden-engineering-unit-test-qualitySkill
- wicked-garden-governed-workerSkill
- wicked-garden-groundSkill
- wicked-garden-jam-brainstorm-facilitatorSkill
- wicked-garden-jam-councilSkill
- wicked-garden-jamSkill
- wicked-garden-learnSkill
- wicked-garden-mcp-scaffoldSkill
- wicked-garden-mem-captureSkill
- wicked-garden-mem-ingestSkill
- wicked-garden-memSkill
- wicked-garden-personaSkill
- wicked-garden-platform-auditSkill
- wicked-garden-platform-compliance-officerSkill
- wicked-garden-platform-complianceSkill
- wicked-garden-platform-errorsSkill
- wicked-garden-platform-gate-benchmark-rebaselineSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 31
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 …
- Distinctive: line 82
Only through the broker: Never add an MCP server to a CLI's configuration, never start an MCP server yourself, and never call an upstream …
- Very specific: line 86
The token is a credential: Never print WICKED_MCP_TOKEN, put it on a command line, or paste it into output, a file or a commit.
≈ “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.
Files
MCP tools through the broker
A governed unit reaches MCP tools in exactly one way: this shim, which sends your unit's capability token and the request to the wicked-crew broker. The broker holds every server connection and every secret, judges the call against the steering policies for your unit (phase role, seat, run mode), runs it, scrubs the result and records the call. You never see a secret, and you never configure an MCP server.
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.
1. List what your unit may try
wicked-garden run scripts/mcp/shim.py listIt prints one JSON object: unit (run, phase, seat) and tools. Each tool has its
subject (mcp:<server>/<tool>), its class (read, write or destructive), its
description, its inputSchema and a decision:
allow: a call runs now.ask: a call waits for the operator's approval (a server's first use, a new schema, or a write the run's mode gates). Calling it records the request for the operator; it does not run.
A tool your unit may never call is not listed: an evaluator or recon unit sees no write or destructive tool, and a tool a policy denies is left out. An empty list means there is nothing for this unit; continue without MCP tools.
2. Call one tool
wicked-garden run scripts/mcp/shim.py call mcp:<server>/<tool> --args '{"key": "value"}'- Build the arguments from the tool's
inputSchema. Put them in a file and pass--args-file <path>(or--args-file -for stdin) when quoting is awkward. - The
mcp:prefix may be omitted (<server>/<tool>). - It prints one JSON object, the broker's answer, with an
outcome. Onokthe tool's result is inresult. The result can itself be a tool error (result.isError): read it and fix your arguments, do not assume success.
3. Read the answer by its exit code
| Exit | Outcome | What to do |
0 | ok | Use result. Cite the call (subject, callId) where you rely on it. |
3 | denied or withheld | Blocked by policy (ruleIds, reason, remedy). Final: do not retry it with other arguments or another route. Continue without it and say so in your output. |
4 | pending_approval | The operator has not approved it yet. Do not retry in a loop. Do other work, and name the subject in your output so the operator can approve it. |
5 | budget_exhausted, breaker_open, upstream_error | The tool is unavailable for now. A write may or may not have applied after upstream_error: check before repeating it. |
2 | no_mcp_channel, bad_request | Not a governed unit (no WICKED_MCP_TOKEN or WICKED_CREW_URL), or bad arguments. Fix the arguments; without a channel, continue without MCP tools. |
1 | anything else | An invalid token, a guard error or an unreachable broker. Report it; do not work around it. |
A denied call never fails your unit; only you can make it look like it succeeded, so report it honestly.
Rules
- Only through the broker. Never add an MCP server to a CLI's configuration, never start an MCP server yourself, and never call an upstream API directly to get around a refusal. A native MCP tool your CLI happens to offer is judged the same way and an unregistered one is denied.
- The token is a credential. Never print
WICKED_MCP_TOKEN, put it on a command line, or paste it into output, a file or a commit. The shim never prints it. - Evaluators read, creators write. In a review, verify or recon unit only read tools are offered and a write is denied by the engine, whatever the run mode.
- Say what you called. List each MCP call you relied on (subject, outcome,
callId) in your output; the evaluator and the run's Governance panel read the same record.
Manual alternative (no launcher or no Python): send the same request with curl,
passing the body on stdin so the token never appears on a command line:
curl -s -X POST "$WICKED_CREW_URL/api/v1/mcp/tools" -H 'Content-Type: application/json' --data-binary @- <<BODY
{"token": "$WICKED_MCP_TOKEN"}
BODY
curl -s -X POST "$WICKED_CREW_URL/api/v1/mcp/call" -H 'Content-Type: application/json' --data-binary @- <<BODY
{"token": "$WICKED_MCP_TOKEN", "subject": "mcp:<server>/<tool>", "args": {}}
BODYThe answers are the JSON objects described above; the HTTP status stands in for the exit code (200 ok, 403 denied or withheld, 409 pending approval, 429, 502, 503 or 504 unavailable, 401 invalid token).
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
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.

