Idea Genie

One canonical root for idea work: elicit an evidence-grounded portfolio of options, or challenge a consequential idea with sealed independent perspectives. Both modes explore and advise; neither selects, schedules, tracks, implements, or validates work.

Modes

Elicit is the entry mode. Duel is an optional escalation for a consequential choice, typically consuming an idea-portfolio.v1 or a framed question. A scored multi-member duel, where members score each other's ideas, is Council's duel mode; a plan's failure modes belong to Premortem.

Both validators live in this skill's scripts/ directory and need jq. Without jq, check the fields by hand against the shape and say the validator did not run.

<skill-dir> is the directory that contains this SKILL.md. Claude Code shows it as "Base directory for this skill"; Codex lists this skill's SKILL.md path; in an AgentOps source checkout it is skills/idea-genie. Replace <skill-dir> with that absolute path before running a command. The working directory stays the user's repository. If the host did not show where this skill is installed, use the directory of the SKILL.md you read. If you cannot determine it, do not guess a path or run a same-named script from the current repository; say the bundled script is unavailable and check the fields by hand against the shape, as without jq.

Elicit mode

These rules carry the value:

1. Observations cite a source: a file and line, issue, doc or measurement. Anything uncited is an assumption, listed apart and never used as support.
2. Every candidate cites the observations behind it. No cited support, no candidate; an unsupported idea may stay listed as an assumption. Never pad the list to a count.
3. Check overlap before claiming novelty. Compare each candidate with what the product already does and with earlier decisions that still apply (an ADR, closed issue or recorded lesson). A request an existing capability already covers is not new: record it under overlaps of the candidate it sharpens, or leave it out. Keep an idea a decision rejected only with new evidence that answers its reason, citing the decision under overlaps.
4. One Given/When/Then per candidate, a normal or edge case whose Then shows the cited problem easing; a Then that could hold while the problem remains aims at a proxy. A goal or a symptom is not yet the problem: cite its cause, or list the cause as an assumption and let the Then show the cited symptom easing.
5. Do not rank, pick, schedule or start work. The caller or Plan selects.
6. Stop at saturation. Merge equivalents, and run another pass only while it adds a materially new evidenced candidate. Zero candidates is valid.
7. Generate before judging. Split the subject into the parts its evidence shows, named in its own words rather than as a directory listing, and draft across each before filtering. Then reshape the predictable first ideas: remove or invert a step, break an assumption treated as fixed, push a constraint to an extreme, borrow how another field solves the same problem, or find a move that makes later moves cheaper. A candidate any project could list unchanged gets a specific observation or is dropped. Keep a one-line reason for each dropped idea in dropped so a later pass does not regenerate it, and list there any part left with no supported candidate.
8. Behavior outweighs stated interest. "Users asked for it" is weaker than what they do now: the workaround they already use, what it costs them, and observed adoption of anything similar. Name the status quo each candidate must beat in status_quo; if nobody is working around the problem today, say so there.

State the question, constraints, non-goals and sources first. Read only the sources this question needs and cite each one; do not build a merged context store.

Inside a named part (a flow, command, page or stage), go as far as the evidence allows: a candidate that reshapes the wider product is a scope overrun, and one that swaps the subject for another domain replaces it. Drop either with that reason unless a cited observation shows the part cannot change without it. With no named subject, sample each top-level area and the recent commits instead of guessing one focus, and let each candidate's evidence show why its subject deserves attention.

Return the portfolio in this shape:

{
  "schema_version": "idea-portfolio.v1",
  "status": "candidates",
  "observations": [{"claim": "<what is true>", "evidence": "<cited source>"}],
  "assumptions": ["<belief no cited source supports>"],
  "candidates": [{
    "id": "I1",
    "idea": "<short name>",
    "evidence": ["<cited source>"],
    "overlaps": ["<existing capability or decision it extends or repeats>"],
    "status_quo": "<what people do today instead, or none>",
    "scenario": {"given": "<state>", "when": "<event>", "then": "<observable result>"}
  }],
  "dropped": [{"idea": "<short name>", "reason": "<overlap, no evidence, or why it lost>"}]
}

Each candidate evidence entry repeats an observation's evidence exactly. Candidate ids are unique and overlaps may be empty. dropped is optional while there are candidates. When every idea overlaps or lacks support, set status to no-new-work, leave candidates empty and list each idea in dropped with the capability, decision or missing evidence that removed it. observations still needs one entry or more: cite what you read, even when it supports no candidate. The validator rejects any other field, so no rank or winner rides along. For a person, render the same fields as a short list, one block per candidate. When a file is wanted, write .agents/scratch/ideas/<run-id>/idea-portfolio.json and check it with bash "<skill-dir>/scripts/validate-output.sh" <path> before returning it to the caller.

Duel mode

Produce independent challenges for a consequential choice. The result is advisory evidence for Plan: return it to the caller, who may supply it to Plan beside research and operator intent. It never decides whether a plan is ready and never turns a later optional Premortem challenge into an approval gate. Emit no readiness, approval, quorum, retry, budget, helper, delivery, or tracker state: this strategy supplies evidence, not lifecycle authority. Consensus, transport availability or a self-score never becomes readiness.

Door class. Ask what undoing the choice after it lands would cost. A one-way door needs a migration, breaks a published contract or caller, loses data, or has an external effect that cannot be recalled: run the workflow below. A cheap two-way door is undone by a revert or a flag: run Plan's one fresh challenge instead and return its findings, with no packet, panel, pane manager or model-family rule.

Workflow

1. Freeze the question, constraints, evidence paths and comparison rubric. Leave unknowns open. An assumption only one approach needs stays with that approach; in the shared constraints it would rule rivals out unseen.
2. Start at least two fresh contexts: separate subagents or sessions with no shared transcript, one prompt each carrying the frozen question and evidence paths and asking for the return in sealed perspectives. Record each native context id as context_id, and collect every perspective before revealing any, so later proposals cannot anchor on earlier ones. If you cannot start two fresh contexts, say so, write no packet, and return the open question with the check that would settle it; never play a perspective yourself or make up a context id. When the caller pins perspectives to model profiles, dispatch them as model-dispatch describes and record each profile name as model_identity; disclose an unavailable profile and continue single-model.
3. Reveal the sealed perspectives and cross-review every one on all five of evidence, reversibility, system fit, failure modes and cost. Each review comes from the reviewing perspective's own context, or from a fresh one given only the proposals once that context has closed; never write a review under a perspective id yourself. Compare perspectives by mechanism, not by name: two that rely on the same mechanism count as one alternative, and fewer than two distinct ones is an incomplete comparison, not a winner; one bounded recovery is in convergence. A perspective that breaks a frozen constraint is out, whatever its other strengths. A combination of parts from several perspectives is a new candidate with no inherited support: walk one concrete example through each guarantee the combined design needs before listing it, or list it as unchecked so Plan does not combine on trust.
4. Attempt concrete refutations. A premise of the frozen question that every perspective builds on gains no support from their agreement: check it at its source. Refute a stateful mechanism with a state trace, not prose. Keep disagreements, failed refutations and minority reasoning explicit, so the caller sees the alternatives a synthesis would erase.
5. Return the packet below. When a file is wanted, write .agents/scratch/ideas/<run-id>/idea-challenge.json and check it with bash "<skill-dir>/scripts/validate-challenge.sh" <path>.

{
  "schema_version": "idea-challenge.v1",
  "door_class": "one-way",
  "sealed_generation": true,
  "perspectives": [
    {"id": "P1", "context_id": "<native context id>", "proposal": "<mechanism, evidence, tradeoff>"},
    {"id": "P2", "context_id": "<a different context id>", "proposal": "<mechanism, evidence, tradeoff>"}
  ],
  "cross_reviews": [
    {"reviewer": "P1", "subject": "P2", "dimensions": {"evidence": "<finding>", "reversibility": "<finding>", "system_fit": "<finding>", "failure_modes": "<finding>", "cost": "<finding>"}},
    {"reviewer": "P2", "subject": "P1", "dimensions": {"evidence": "<finding>", "reversibility": "<finding>", "system_fit": "<finding>", "failure_modes": "<finding>", "cost": "<finding>"}}
  ],
  "disagreements": ["<where the perspectives still differ>"],
  "refutations": [{"claim": "<what was attacked>", "attempt": "<the concrete test>", "result": "<what the attempt showed>"}]
}

Each perspective has a distinct id and context_id and a proposal, plus model_identity when pinned. Every perspective is the subject of at least one cross-review by another; dimensions holds all five rubric keys and no others, each a text finding. disagreements and refutations each need at least one entry. The validator rejects any other field except an optional legacy handoff (owner plan, artifact_dir, optional route), which new packets omit.

References

• Idea Genie behavior
• Idea challenge behavior
• Sealed perspectives