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

wicked-garden-jam

Skill9 starsChanged 7 months agoMITView on GitHub

Orchestrates AI-powered brainstorming sessions with dynamic focus groups. quick sessions are ephemeral (no storage). brainstorm and council sessions are tracked as native tasks (process) and stored via wicked-garden-mem (outcome). Use when: "brainstorm this", "explore ideas", "get different perspectives", "focus group", "what do you think about", "pros and cons", "quick check", "jam quick <idea>", "jam brainstorm <topic>", "jam council <topic> with options A/B/C", "multi-model evaluation", "council verdict", "jam revisit <decision>", "how did that decision work out".

  • Installs packages

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are 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.

  • 3

    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 26
  2. Very specific:

    Read refs/quick.md (relative to this skill's base directory) — the single-pass rubric: 4 personas, 1 forced round, synthesis format (Key …

    line 67
  3. Very specific:

    After the council returns: read refs/council-verdict.md — it holds the caller-side heuristics for acting on the verdict (when to proceed, …

    line 114
  4. Very specific:

    Read refs/revisit.md — the 5-step workflow: recall the decision via the wicked-garden-mem skill, display the decision summary, ask …

    line 128
  5. Distinctive:

    Follow it step by step, waiting for the user's outcome answer in step 3.

    line 133

≈ “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

Brainstorming Skill

Generate diverse perspectives through structured focus group sessions. This is the single entry point for the jam domain — route the user's request to one of the sub-actions below.

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.

Routing

Sub-actionWhenHow it runs
quickGut-check, rapid exploration (~60s)Inline — apply refs/quick.md directly
brainstormImportant decisions, complex problemsHand-off → wicked-garden-jam-brainstorm-facilitator
councilDefined options needing a rigid verdictHand-off → wicked-garden-jam-council
revisitRecord the outcome of a past decisionInline — follow refs/revisit.md

Progression: quick (60s gut-check, ephemeral) → brainstorm (full session with evidence + decision storage) → council (structured verdict with external LLMs, final step). Key distinction: quick = ideation, brainstorm = free-form exploration, council = rigid evaluation of defined options.

Quick-Start

# Fast gut-check (ephemeral, ~60s)
jam quick "Should we use feature flags or config files here?"

# Full session with evidence and decision storage
jam brainstorm "Architecture approach for the new event bus"

# High-stakes council with external LLM challenge
jam council "Go/no-go on the v9 storage migration" --options "go, no-go"

# Record how a past decision worked out
jam revisit "event bus architecture"

Sub-action: quick

Args: <idea or question>

Quick 60-second exploration with 4 personas and 1 round. Run it inline — no dispatch:

  1. Read refs/quick.md (relative to this skill's base directory) — the single-pass rubric: 4 personas, 1 forced round, synthesis format (Key Insights / Action Items / Open Questions), hard constraints (no storage, no multi-AI, ≤200 words).
  2. Apply the rubric directly to the topic. Do NOT run additional rounds. Do NOT store.

Sub-action: brainstorm

Args: <topic> [--personas list] [--rounds n] [--converge fast]

Full brainstorm with evidence-backed perspectives. The facilitator gathers evidence from the ecosystem (past decisions, code context, brainstorm outcomes) before assembling personas, so they argue from data — not just opinions. After synthesis, a structured decision record is automatically stored via the wicked-garden-mem skill for organizational memory.

Convergence modes (pass through as convergence_mode):

  • normal (default): Run all requested rounds (2-3), then synthesize.
  • fast (--converge fast): After each round, assess whether there is enough signal to synthesize; skip remaining rounds when personas converge.

Hand off to the facilitator worker skill (it owns the convergence checks, task tracking, transcript storage, and bus events):

Hand-off — open the wicked-garden-jam-brainstorm-facilitator skill with "Run a full brainstorm session on: {topic}. Options: {personas}, {rounds}, convergence_mode={converge|'normal'}." 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.

Sub-action: council

Args: <topic> --options "A, B, C" [--criteria "perf, cost, risk"]

Structured evaluation tool that uses real external LLM CLIs — registry-driven (20+ CLIs: Codex, Gemini, Copilot, OpenCode, Pi, Aider, Goose, Amp, Droid, …; see scripts/jam/agentic_cli_registry.py) — to get genuinely independent model perspectives. Installed CLIs are detected AND usability-probed via wicked-garden run scripts/jam/detect_clis.py --probe (auth-revoked / unconfigured / daemon-down CLIs are excluded). When fewer than 2 usable external CLIs are present, council seats are filled with separate-worker seats so deliberation always happens. Unlike brainstorm (free-form creative exploration), council is a rigid evaluation tool for when you have defined options and need a verdict.

Natural workflow: brainstorm → identify candidates → council → decide

Hand off to the council worker skill:

Hand-off — open the wicked-garden-jam-council skill with "Run a council evaluation on: {topic}. Options: {options}. Criteria: {criteria}." 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.

After the council returns: read refs/council-verdict.md — it holds the caller-side heuristics for acting on the verdict (when to proceed, when to surface raw votes and pause for human adjudication, hard-gate archetype rules) and the raw_votes output envelope contract (consensus.py::build_council_output, WG_COUNCIL_OUTPUT=both|synth|raw).

Sub-action: revisit

Args: <topic or decision keyword>

Revisit a past brainstorm decision to record whether it was validated, invalidated, or modified. Light workflow — run it inline, no hand-off:

  1. Read refs/revisit.md — the 5-step workflow: recall the decision via the wicked-garden-mem skill, display the decision summary, ask validated/invalidated/modified, store the outcome (about tags jam,outcome), report. Degrades gracefully when the memory layer is absent.
  2. Follow it step by step, waiting for the user's outcome answer in step 3.

Workers

  • skills/jam-brainstorm-facilitator/SKILL.md — multi-round, evidence gathering, transcript storage, decision record
  • skills/jam-council/SKILL.md — registry-driven multi-model council with isolation-enforced parallel seats

(jam quick runs inline via refs/quick.md — the former quick-facilitator agent was retired; the ref is the sole, up-to-date rubric.)

References

  • refs/quick.md — the single-pass quick-jam rubric (personas, round, synthesis shape, hard constraints)
  • refs/revisit.md — the decision-outcome revisit workflow
  • refs/council-verdict.md — caller-side verdict heuristics + raw_votes output contract
  • refs/facilitation-patterns.md — persona archetype pool, session length guidance, anti-patterns
  • refs/synthesis-patterns.md — synthesis structure, quality checklist, decision record format

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.