vibe-brainstorm
Use when the user explicitly asks for vibe brainstorming, creative implementation ideas, implicit expected behavior, or convention checks, and when an implementation task is creative or convention-dependent. Do not use for obvious mechanical edits.
How do I install this agent skill?
npx skills add https://github.com/adhi-jp/agent-skills --skill vibe-brainstormIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The 'vibe-brainstorm' skill is a benign, read-only brainstorming framework. It provides structured prompts and guidelines for creative ideation, convention checking, and model selection. It includes strict 'read-only' boundaries, explicitly forbidding unauthorized file edits or command execution, and uses 'Durable Records' for documenting decisions in a transparent way. No malicious patterns were detected.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Vibe Brainstorm
Turn an underspecified creative request into ideas, expected behavior, and a direction or checklist to carry forward. This is a read-only, pre-implementation phase: a trusted proxy may select input for later requirements or planning, but it is not user confirmation or implementation authorization.
Use and mode
Use this skill for requested ideas, alternatives, interaction behavior, or convention-dependent implementation. Skip it for mechanical edits, a direct bug fix with a concrete plan, approved execution, review, or a saved requirements artifact.
| Situation | Mode |
|---|---|
Explicit diverge or ideas only | diverge |
| Autonomous or convention-check request | conventions |
Explicit full, or explicit invocation with a creative goal | full |
Do not autonomously choose full unless ideation, convention grounding, and
selection are all needed to scope the work; when you do choose it, say why.
Perspectives and delegation
Use separate perspectives locally by default; delegate only when independent
work or specialist context materially helps and the host permits it. Never
claim delegation from headings or polished prose. Call it confirmed only
with a citable host record; otherwise say unproven or unavailable.
Delegated prompts request concise candidates, tradeoffs, risks, or checks, not private reasoning. A user-stated scripted orchestration mechanism is sufficient to plan its bounded schedule. If a scripted run is available, it may cover generation, critique, development, grounding, and selection, but final confirmation stays in chat. If the current host cannot actually invoke or record that user-stated run, do not replace it with coordinator-generated brainstorm results. Return the planned stages, role-specific capability-tier basis, required run/task evidence, and the post-run conversation confirmation boundary instead. For ordinary perspective gathering, give local coordinator-derived perspectives unless independence itself is required; then state the limitation. Local coordinator-derived perspectives are the mode's own work and need no delegation or fallback authorization.
Model Choice
<!-- shared-contract:class language=none commit=none effect=read-only --> <!-- shared-contract:begin model-tier-selection source=shared/vibe-contract.md -->Choose a model for each delegated unit by capability and context fit, never by hard-coded name.
- Only choose when the host allows it and the user has not fixed a model.
- Use a cheaper or faster tier only for bounded, low-ambiguity lookups, extraction, and mechanical checks; use the strongest suitable tier for synthesis, adversarial or security review, human-risk reasoning, and final recommendations, or when the user asks for maximum performance.
- Record the choice only for a user override, degraded capability, a cost or performance constraint, or audited external execution.
Creative synthesis, convention tradeoffs, selection, and final recommendations need the stronger suitable capability; cheap tiers fit only bounded checks. In an orchestration schedule, explain each role's capability and context needs; name a tier or model only under one of the block's recording conditions, and name that condition.
Modes
diverge
Generate Practical, Unconventional, and Challenging directions. For each,
give fit, tradeoffs, and implementation risks. Do not ground conventions,
rank, choose, recommend, begin implementation, or offer to implement a
direction.
conventions
Summarize the behavior, then produce a checklist separating mandatory expected behavior, optional taste, and unknowns. Ground it in relevant local, supplied, official, or domain sources when available, and say which items each source grounds; label missing access and inference. Mandatory means missing it would violate the goal, a real convention, accessibility/safety expectation, data contract, or normal user expectation. Surface meaningful UX tradeoffs rather than selecting the cheapest build path. Hand the checklist forward as AI-selected input; stop for user confirmation only when it adds, drops, or changes requested behavior or scope. Spelling out behavior the request implies is not such a change.
full
Generate the three directions, run the convention checklist, develop viable candidates, and apply this sieve: reject failures of mandatory gates (name the gate, or say that none failed), then rank survivors for creativity, fit, and practicality. Recommend a direction and any worthwhile runner-up. Do not reject unusual ideas merely for being unusual. Stop for user confirmation; a proxy-selected direction is only AI-selected input for later requirements or planning.
Output and boundary
Effect And Write Boundaries
<!-- shared-contract:begin effect-write-boundaries source=shared/vibe-contract.md -->Write nothing beyond what the phase's declared effect class and boundary permit.
- Apply the effect class the phase declares in its own text; where this package states a narrower limit, the narrower limit wins.
- Read-only: report in chat; edit no file, run no state-mutating command, and never stage, commit, tag, push, change versions, delete data, or start services. Write a file only when the user explicitly asks for a saved artifact.
- Artifact-only: create or update only the artifact the phase owns and the supporting paths its text declares, such as a confirmed plan reflection or a decision record, and leave them in the working tree.
- Never let an artifact-only phase implement executable behavior, edit code or tests as implementation, produce another phase's artifact, do release work, or treat its artifact as same-turn implementation authority.
- State-changing: edit and run commands only inside the declared scope, in its smallest verified unit; leave other paths, pre-existing changes, and runtime or external state untouched unless the user selects them.
- These limits cover shell commands (redirection,
sed -i,tee,mv,cp,rm,git checkout --) as well as file tools; report a refused write as a boundary stop and never retry it through another tool. - Keep every irreversible or outward-facing operation under its own consent.
Keep the result in chat unless the user explicitly requests a saved artifact.
Response Shape
Report the mode and delegation state, labeling locally produced results coordinator-derived; the checklist and candidates where generated; selection or skipped stages; and the exact checklist or direction, saying whether it awaits user confirmation or goes forward as AI-selected input. Summarize conclusions, evidence, tradeoffs, and open questions, never private reasoning.
Durable Records
Read references/durable-records.md only before handing a settled decision or
deferred finding forward. This phase writes neither decision nor finding
records; its packet remains unpersisted until a later writing phase records it.
Handoff
Stop at the user's pick of a full direction or a proxy selection, and for
user confirmation of a checklist that adds, drops, or changes requested
behavior or scope; hand any other conventions checklist forward as
AI-selected input without waiting.
Hand the outcome forward as a carry-forward packet; without user
confirmation, do not implement, edit, stage, commit, or call it approved.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/adhi-jp/agent-skills/vibe-brainstorm">View vibe-brainstorm on skillZs</a>