agentleFS
BolivarTechBolivarTech/claude-skills/design-brainstorm/SKILL.md

design-brainstorm

Skill0 starsChanged 51 days agoMITView on GitHub

Turn an idea, feature request, or proposed behavior change into an approved design or behavioral specification before any implementation starts. It classifies the request as a feasibility spike, a bounded change, or an architectural change, explores the repository first, presents the design, and stops for approval; when premortem reports or gap registries exist, it carries their risks, controls, and open gaps into the design. Use for new features, new subsystems, interface changes, ambiguous requests, feasibility questions, or when the user says "let's brainstorm this", "design this before we build it", "how should we approach this feature", "diseñemos esto antes de implementar", "quiero agregar una funcionalidad", "cómo encaramos este cambio". Not for implementing an already approved spec or plan, diagnosing a bug, postmortems, or ordinary code review.

2 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions2 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 2

    1. kopp0510/claude-dd★ 12same name
    2. hsiangyilu/claude-design-skills★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository9 other files it gives its agents.

The file

Show its rules and text

We split this skill into the rules it gives an agent

28 rules. About 0 are common, 27 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈27

    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
  1. Distinctive:

    When uncertain, choose the heavier path.

    line 44
  2. Distinctive:

    Use a repository index or knowledge graph first when local rules require it.

    line 57
  3. Distinctive:

    Ask only questions whose answers materially change behavior or scope, and ask one focused question at a time.

    line 60
  4. Distinctive:

    Search for premortem material in the repository's declared process directories and in paths such as:

    line 64
  5. Distinctive:

    Stop and obtain explicit approval.

    line 98
  6. Distinctive:

    After approval, follow the repository's normal implementation and TDD workflow.

    line 99
  7. Distinctive:

    Write the specification at the repository-defined path.

    line 115
  8. Distinctive:

    Ask the user to review and approve the written specification.

    line 123
  9. Distinctive:

    Only after approval, transition to the repository-defined planning workflow.

    line 124
  10. Distinctive:

    Do not commit a design or spec unless the user has explicitly authorized a commit.

    line 126
  11. Distinctive:

    Keep the user's objective and scope fixed.

    line 130
  12. Distinctive:

    Prefer the simplest complete design with a current consumer.

    line 132
  13. Distinctive:

    Do not present mocks, inspection, or old command output as live integration evidence.

    line 142
  14. Distinctive:

    Every path has an approval before action:

    line 148
  15. Distinctive:

    Read-only repository exploration is allowed before approval.

    line 159
  16. Distinctive:

    Use this section only when a premortem report or gap registry matches the brainstorming objective.

    line 165
  17. Distinctive:

    If relevance is uncertain:

    line 174
  18. Distinctive:

    Use findings that remain technically true.

    line 177
  19. Distinctive:

    Do not import conclusions that depend on changed assumptions.

    line 178
  20. Distinctive:

    Read the report and any linked gap registry.

    line 183
  21. Distinctive:

    Do not treat generated causes as verified facts.

    line 195
  22. Very specific:

    When a behavioral baseline (for example SBTDD's spec-behavior-base.md and spec-behavior.md) and premortem artifacts coexist:

    line 215
  23. Distinctive:

    Treat the behavioral baseline as the source contract.

    line 218
  24. Distinctive:

    Use the premortem to find missing boundaries, negative scenarios, controls, and evidence.

    line 219
  25. Distinctive:

    Verify whether premortem corrections are already present in the baseline.

    line 220
  26. Distinctive:

    Add no behavior solely because the premortem imagined it.

    line 223
  27. Distinctive:

    If the premortem records an exact sibling revision, pin, fixture hash, or release artifact, carry that value into the spec.

    line 227
  28. Distinctive:

    Do not rerun or rewrite the premortem merely to make the design look complete.

    line 242

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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

Design Brainstorm

Clarify intent and turn it into the smallest design that fully satisfies the user's objective. Explore the repository before proposing changes. No implementation, scaffolding, production-code edit, or implementation skill begins until the user approves the proposed direction. The amount of design ceremony may shrink; the approval gate does not.

Work in the user's language. Identifiers, paths, commands, and quoted strings stay as they are.

Repository rules take precedence. If they define SDD, BDD, SBTDD, artifact locations, approval checkpoints, or a planning workflow, use those instead of inventing parallel documents.

Classify the request

State the classification briefly so the user can correct it:

  • Spike: A feasibility question whose deliverable is an answer or recommendation, not production code.
  • Bounded: A focused change to an existing flow whose interfaces and consumers can be inspected in the repository.
  • Architectural: A new project, subsystem, execution route, persistent format, shared interface, or change that restructures components or affects several consumers.

When uncertain, choose the heavier path. If repository inspection reveals wider consequences, upgrade the classification and stop before implementation. Do not downgrade once material hidden complexity is known.

Gather context before asking questions

Inspect the smallest useful set of sources:

  1. Repository rules and the active spec, plan, and session state.
  2. Relevant implementation, tests, public interfaces, and recent history.
  3. Existing designs or behavioral baselines for the same objective.
  4. Relevant premortem artifacts, when present.

Use a repository index or knowledge graph first when local rules require it. Prefer scoped search over broad file dumps.

Ask only questions whose answers materially change behavior or scope, and ask one focused question at a time. When the repository already answers a question, make the safe assumption and disclose it instead of asking.

Search for premortem material in the repository's declared process directories and in paths such as:

  • sbtdd/premortem*.md
  • planning/premortem*.md
  • **/premortem-*.md
  • **/premortem-gaps.md

Process artifacts are often gitignored. Search the declared directories directly or include ignored files (for example, rg --files -uu) instead of relying only on the tracked file list.

Apply the premortem integration section below when a relevant report or gap registry exists. Ignore unrelated premortems. Never treat a filename match as proof that a report covers the current objective.

Choose the path

Spike

  1. Inspect enough context to frame the uncertainty.
  2. Present the question, success signal, probe, limits, and what would remain unknown in two or three sentences.
  3. Obtain approval.
  4. Investigate as cheaply as correctness allows.
  5. Report evidence and a recommendation. Label any temporary artifact as throwaway; keeping it requires a separately classified change.

Bounded

  1. Inspect the existing flow, callers, tests, and constraints.
  2. Resolve the questions that remain after inspection, one at a time.
  3. Present a short design in chat: behavior, affected interfaces/files, failure handling, compatibility, and verification.
  4. Identify premortem risks or open gap IDs that constrain the change.
  5. Stop and obtain explicit approval.
  6. After approval, follow the repository's normal implementation and TDD workflow. Do not create a separate plan artifact unless repository rules require one.

Architectural

  1. Establish objective, actors, observable behavior, exclusions, constraints, compatibility, success criteria, and unresolved decisions.
  2. Decompose the request if it contains independent subsystems that cannot be reviewed or verified as one bounded contract.
  3. Present two or three viable approaches with trade-offs. Lead with the recommended option and explain why it best fits current consumers and constraints.
  4. Present the design in reviewable sections. Cover boundaries, data/control flow, errors, cleanup, security, compatibility, migration, testing, and operational evidence when relevant.
  5. Reconcile applicable premortem findings. Carry open gap IDs into the design rather than claiming they are closed by prose.
  6. Obtain approval of the design direction.
  7. Write the specification at the repository-defined path. When no path is defined, propose one and confirm it with the user before writing. If an approved behavioral baseline exists, preserve its requirement and scenario IDs and produce the refined spec required by the local workflow. If a specification already exists at that path, review it against the baseline, the repository, and the gap registry instead of writing a second one; deliver the findings and keep both gates.
  8. Self-review the written spec and fix unsupported assumptions, placeholders, contradictions, ambiguous behavior, missing negative cases, and traceability gaps.
  9. Ask the user to review and approve the written specification.
  10. Only after approval, transition to the repository-defined planning workflow.

Do not commit a design or spec unless the user has explicitly authorized a commit.

Design discipline

  • Keep the user's objective and scope fixed. A risk control may constrain an implementation; it does not silently create a new product feature.
  • Prefer the simplest complete design with a current consumer.
  • Reuse compatible repository and sibling implementations before proposing new machinery.
  • Separate observable behavior from implementation choices. Put behavior in the spec and design choices in the plan unless the choice is externally visible.
  • Distinguish verified facts, inherited requirements, assumptions, decisions, and unresolved evidence.
  • Preserve exact identifiers, versions, hashes, paths, flags, error strings, and claim strength.
  • Include boundaries for empty and malformed input, unavailable dependencies, timeout, cancellation, cleanup, concurrency, security, and compatibility when they can affect the objective.
  • Do not present mocks, inspection, or old command output as live integration evidence.
  • Re-verify evidence recorded by an earlier artifact against the current tree before the design relies on it, and record the date or revision of the check.

Approval gates

Every path has an approval before action:

  • Spike: approve the probe before investigation that creates artifacts or spends external resources.
  • Bounded: approve the short design before implementation.
  • Architectural: approve the design direction before writing the final spec, then approve the written spec before planning.

At each gate, list the decisions that only the user can make, with a recommendation for each; approval is not requested until they are answered or explicitly deferred.

Read-only repository exploration is allowed before approval. Approval for brainstorming or a spec does not authorize implementation, commits, remote changes, releases, or other external mutations.

Premortem integration

Use this section only when a premortem report or gap registry matches the brainstorming objective.

Determine relevance

A report is relevant when its subject, target artifact, version, and decision match the current request. Check its date or source revision and whether later rule, scope, backend, interface, or baseline changes invalidate it.

If relevance is uncertain:

  • State what matches and what may be stale.
  • Use findings that remain technically true.
  • Do not import conclusions that depend on changed assumptions.
  • Recommend a new premortem only when the current trigger policy requires one or the user asks.

Extract useful evidence

Read the report and any linked gap registry. Capture:

  1. The imagined failure and success condition.
  2. Prioritized failure causes and their evidence.
  3. Preventive controls, early signals, contingencies, and owners when recorded.
  4. Changes already applied to the baseline or design.
  5. Accepted risks, if the acceptance authority is explicit.
  6. Open implementation or external-evidence gaps and their stable IDs.
  7. Assumptions, unknowns, handoffs, and review triggers.
  8. The "second barrel" or status-quo alternative, including reversibility and the cost of not acting.

Do not treat generated causes as verified facts. Preserve their stated confidence and evidence.

Reconcile with the design

For every relevant finding, classify it as:

  • Already covered: cite the requirement, scenario, design section, or existing control.
  • Clarification: tighten an existing behavior without changing scope.
  • Planning obligation: carry it into a task, gate, owner, or dependency.
  • Implementation gap: preserve its gap ID and required closure evidence.
  • External evidence: name the real backend, platform, account, reviewer, or release artifact needed to close it.
  • Out of scope: explain why it does not apply to the approved objective.
  • Scope expansion: present it separately and require user approval before adding it.

Prose does not close an implementation or external-evidence gap. Update a gap status only after the cited evidence exists. A closed planning ambiguity may remain an open runtime risk.

Use with a behavioral baseline

When a behavioral baseline (for example SBTDD's spec-behavior-base.md and spec-behavior.md) and premortem artifacts coexist:

  1. Treat the behavioral baseline as the source contract.
  2. Use the premortem to find missing boundaries, negative scenarios, controls, and evidence.
  3. Verify whether premortem corrections are already present in the baseline.
  4. Inspect the repository and record evidence for each refinement.
  5. Preserve every baseline requirement and scenario ID in the refined spec.
  6. Add no behavior solely because the premortem imagined it; obtain approval for new scope.
  7. Include a reconciliation table from requirements and scenarios to repository evidence and open gap IDs.

If the premortem records an exact sibling revision, pin, fixture hash, or release artifact, carry that value into the spec. If it records only "latest" or a mutable branch, resolve and record an immutable reference before treating it as reproducible.

Required outcome

The design or spec must make clear:

  • Which premortem findings changed or constrained the design.
  • Which findings were already covered.
  • Which gap IDs remain open and what evidence closes them.
  • Which assumptions still need user approval or runtime proof.
  • Why the proposed path is preferable to the status quo when the report contains a second-barrel analysis.

Do not rerun or rewrite the premortem merely to make the design look complete. Preserve the report as analysis and use the gap registry as the live closure interface.

Output

Scale the result to the path:

  • Spike: question, probe, findings, remaining uncertainty, recommendation.
  • Bounded: concise in-chat design and approval request.
  • Architectural, at the direction gate: classification, verified evidence, approaches with the recommendation, premortem reconciliation, and the user decisions.
  • Architectural, after approval: repository-defined Markdown specification with requirements, scenarios or acceptance examples, constraints, exclusions, evidence, assumptions, decisions, premortem reconciliation, and traceability.

The final artifact must say what is approved, what remains open, and what the next authorized workflow step is.

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.