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

wicked-garden-agentic

Skill9 starsChanged 7 months agoMITView on GitHub

Agentic-systems domain skill: review, design, audit, and framework selection for AI agent systems. Routes to one of four actions (review | design | audit | frameworks) backed by the rubrics in refs/.

  • Installs packages

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 1 is distinctive, and 8 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.

  • 8

    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 32
  2. Very specific:

    Read refs/review.md (relative to this skill's base directory) — the full 5-step review rubric: framework+topology detection, architecture …

    line 57
  3. Very specific:

    Run the detection scripts from the ref (Step 1).

    line 61
  4. Very specific:

    Write to --output file when set.

    line 63
  5. Very specific:

    Read refs/design.md — the design rubric: requirements gathering, pattern selection, five-layer architecture, safety section, framework …

    line 86
  6. Very specific:

    Read refs/audit.md — the 8-layer rubric, checklist, compliance extensions, and output format.

    line 100
  7. Very specific:

    If --standard is given, append the matching compliance checklist from the ref.

    line 104
  8. Very specific:

    If --scenarios, emit a wicked-scenarios block per CRITICAL/HIGH finding.

    line 105
  9. Very specific:

    Use WebSearch for latest 2026 ecosystem state (versions, features, community) before rendering the comparison or recommendation.

    line 119

≈ “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

Agentic Domain

One entry point for the agentic domain. Pick the action from the request, parse its args, load its ref, and apply it inline.

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.

Action router

ActionUse forArgsRef
reviewFull agentic-codebase review: framework detection, topology, architecture + safety + performance, remediation roadmap[path] [--quick] [--framework NAME] [--output FILE]refs/review.md
designGreenfield agentic system design: requirements → pattern + five-layer architecture → safety validation → design doc[problem description] [--output FILE]refs/design.md
auditCompliance-grade trust+safety audit: tool risk classification, HITL gates, PII handling, evidence[path] [--standard GDPR|HIPAA|SOC2|NIST] [--output FILE] [--scenarios]refs/audit.md
frameworksFramework selection / comparison / wizard[--compare fw1,fw2,...] [--language python|typescript|java|go] [--use-case TYPE]refs/frameworks.md

Routing hints:

  • Assessing existing agentic code → review. Compliance-grade safety evidence → audit. Greenfield design → design. Picking or comparing frameworks → frameworks.
  • Ordinary (non-agent) source code → engineering review, not this skill.
  • Binding go/no-go verdict → archetype review, not this skill.

Action: review

Full agentic-codebase review: framework detection → topology → architecture + safety + performance assessments → pattern scoring → unified remediation roadmap.

  1. Parse [path], --quick, --framework, --output.
  2. Read refs/review.md (relative to this skill's base directory) — the full 5-step review rubric: framework+topology detection, architecture assessment, safety 8-layer, performance assessment, pattern scoring + issue taxonomy, and output format.
  3. Run the detection scripts from the ref (Step 1). If --quick, stop and return the structural summary. Otherwise apply all rubric steps directly.
  4. Write to --output file when set; otherwise return inline.

Heavyweight full review (optional hand-off)

For large codebases, or when independent parallel assessments are wanted, hand off to the three worker skills — in parallel where your harness runs parallel sub-agents, serially otherwise (record serial_reason: harness has no parallel dispatch — the sanctioned degradation) — instead of applying rubric steps 2–4 inline, then merge their findings into the ref's unified roadmap:

  • wicked-garden-agentic-architect — five-layer architecture + agent topology
  • wicked-garden-agentic-safety-reviewer — guardrails, prompt injection, PII, HITL
  • wicked-garden-agentic-performance-analyst — tokens, latency, cost, parallelization

Hand-off — open each worker skill by name with the target path and the --quick flag as the argument; on Claude Code this is the Skill tool, on any other seat open the named skill from your catalog and carry it out inline, then continue here.

Action: design

Interactive design session for a new agentic system: requirements → pattern + five-layer architecture → safety validation → design doc. Greenfield only — use review to assess existing code, audit for compliance evidence.

  1. Parse [problem description] and --output.
  2. Read refs/design.md — the design rubric: requirements gathering, pattern selection, five-layer architecture, safety section, framework recommendation, and output format.
  3. Work through the rubric phases directly. If no problem statement is supplied, ask the 3–5 clarifying questions from the ref before proceeding.
  4. Write to --output file when set; otherwise return inline.

Action: audit

Deep trust+safety audit: classifies tool risks, verifies HITL gates, checks PII handling, optionally emits compliance evidence + wicked-scenarios. Use review for the broader architecture+perf+safety sweep.

  1. Parse [path], --standard, --output, --scenarios.
  2. Read refs/audit.md — the 8-layer rubric, checklist, compliance extensions, and output format.
  3. Apply the rubric directly to the target path. For each layer, assess findings, classify severity, and build the risk matrix.
  4. If --standard is given, append the matching compliance checklist from the ref.
  5. If --scenarios, emit a wicked-scenarios block per CRITICAL/HIGH finding.
  6. Write to --output file when set; otherwise return inline.

Action: frameworks

Framework selection / comparison / wizard. NOT for reviewing existing agentic code (use review) or architecture design (use design).

  1. Derive mode from args: --compare → side-by-side comparison; filters only (--language, --use-case) → filtered selection; no args → interactive 5-question wizard.
  2. Read refs/frameworks.md — the mode detection table, wizard questions, decision tree, comparison table, scoring template, and output format.
  3. Use WebSearch for latest 2026 ecosystem state (versions, features, community) before rendering the comparison or recommendation. The curated framework profiles live in the skills/agentic/frameworks/ knowledge module.
  4. End output with a pointer to the design action of this skill as the next step.

Knowledge modules

The domain's reference knowledge lives beside this skill and is loaded on demand:

  • skills/agentic/agentic-patterns/ — pattern catalog + five-layer model
  • skills/agentic/context-engineering/ — context/token optimization techniques
  • skills/agentic/frameworks/ — curated framework profiles + decision tree
  • skills/agentic/review-methodology/ — systematic review approach
  • skills/agentic/trust-and-safety/ — guardrail + HITL patterns

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.