Skip to content
Development
Skill

/whiplash

Critically review a work product, fix supported material issues, and repeat verification and review until no material issue remains; or, as a sweep, apply proven behavior-preserving cleanup across a codebase. Use only when explicitly invoked.

From plugin
hope
76 skills

Claude Code

$ npx -y skills add dkstm95/hope --skill whiplash --agent claude-code

SKILL.md

whiplash.SKILL.md
name: whiplash
description: Critically review a work product, fix supported material issues, and repeat verification and review until no material issue remains; or, as a sweep, apply proven behavior-preserving cleanup across a codebase. Use only when explicitly invoked.
disable-model-invocation: true

Hope Whiplash

Use only when the person invokes `$hope:whiplash`, `/hope:whiplash`, or the host's equivalent skill command or picker. Continue follow-ups within that invoked task without requiring another invocation. Ordinary requests and calls from another skill do not activate this skill; continue authorized work through the ordinary workflow. Reading a shared reference does not invoke its skill.

Critically examine whether the work achieves the person's intended state after the change in real use, fix supported material issues, and review the result again. Be strict about the work and respectful toward people.

Read `../write/references/writing-standard.md` for user-facing language.

Scope

Use the named target or the active work established in the conversation. Read the original request and subsequent changes to distinguish which purposes, requirements, and contracts the latest request replaces and which still apply. Earlier agreements do not override a request to change the purpose itself. The authored diff is evidence to review for errors and omissions, not a requirement or the boundary of the work.

Follow affected behavior through existing implementation, tests, contracts, documentation, and call paths. Include related parts that conflict with or prevent the intended state, even outside the diff or initial change list. Extend scope as needed to complete the change; leave unrelated cleanup out. Invocation authorizes reversible local edits within that scope. Carry forward existing authority for other actions; invocation alone does not authorize commits, pushes, publication, or changes to external systems.

Honor a review-only request by reporting checked findings without edits. Ask only when missing information or a consequential decision changes the result, or an action needs authority not already given. Continue independent authorized work while an answer is pending.

Review, fix, and repeat

Use the conversation, relevant sources, and the person's constraints to choose the review's focus, depth, and format. Do not fill a fixed checklist, require a finding count, or impose a report template.

Test important claims in concrete situations, including design choices that may prevent the intended result. Do not change the person's intent to fit the existing implementation. Apply the same evidence standard to your findings, prior conclusions, and user criticism; seek evidence that could overturn each. Distinguish confirmed defects, design choices, and unverified claims. Do not reverse a conclusion without checking or invent a defect to agree with a criticism. Reject unsupported findings with reasons; reopen settled decisions when new requests or evidence change their basis.

When a test fails or a contract conflicts, use the requests and evidence to decide whether the implementation is defective, the expectation is obsolete, or a requirement that still applies has been violated. Preserve valid checks and update obsolete expectations. Neither weakening the intended change to pass tests nor deleting valid verification to hide failures is a repair. Distinguish records, permissions, and failure conditions that must survive from past execution mechanisms that the request replaces.

Fix supported issues in coherent batches and verify affected behavior with appropriate checks and required project verification. Derive checks from the requested outcome and retained requirements; code and documents or tests edited alongside it cannot alone establish success. After each batch, review the latest result for fulfillment of changed requirements, preservation of retained requirements, and consistency across affected parts, including omissions and regressions outside the diff.

Use independent reviewers when their perspective is worth the time or the person requests them. Give fresh contexts the original and subsequent requests, still-valid constraints, and relevant artifacts and evidence. Label the author's interpretations and assumptions rather than passing them as established facts. Judge findings by evidence, not reviewer agreement. Disclose any limit on requested independence.

Sweep mode

When the person asks Whiplash for a sweep, or for cleanup rather than a review of a change, the intended state is a smaller codebase with unchanged public behavior. Use the named repository or the current one, and the whole repository unless the person narrows the scope. Record the revision and working-tree state and preserve unrelated edits.

Establish current behavior from running code and configuration, then remove dead code, duplication, needless branches or wrappers, misplaced abstractions, and repeated work. Share logic only when behavior, ownership, and reasons to change match. Prove a removal has no remaining consumer, including external or dynamic use; a missing text reference alone does not prove safety. Remove dedicated tests, documentation, generation, configuration, and assets with their obsolete consumer, and retain support for remaining paths.

Leave bugs, public-contract or behavior changes, product and compatibility decisions, migrations, dependency changes, and uncertain removals outside the sweep, and do not turn them into unsolicited findings. Verify affected consumers with the narrowest useful checks plus required project verification, and correct or revert regressions the sweep introduced without repairing pre-existing failures. If no proven cleanup exists, say so.

Finish

Continue under existing decisions and authority until the final result has been checked against the intended state, no material issue remains in the affected scope, and required verification is complet

Read more
@dkstm95
Ships withhope

Hope helps people take on more complex work with AI while preserving their judgment and agency.

Stars on GitHub

Momentum

stars in total
7
Star history of dkstm95/hope on GitHub, a chart by star-history.com

Other skills on hope.

align

align

Resolve consequential choices and shared intent before implementation. Use only when explicitly invoked.

@dkstm95@dkstm95Install Skill
design

design

Create or refine interface mockups from goals, references, and feedback. Use only when explicitly invoked.

@dkstm95@dkstm95Install Skill
diff

diff

Explain or review a GitHub pull request or local Git changes with captured evidence and an offline HTML review with grouped code.…

@dkstm95@dkstm95Install Skill
pr-writing

pr-writing

Write or update PR and MR titles and descriptions when creating or editing a pull request or merge request, including when the…

@dkstm95@dkstm95Install Skill
write

write

Draft, edit, or review language while preserving meaning and voice. Use when explicitly invoked or when Hope PR Writing calls…

@dkstm95@dkstm95Install Skill

More skills from other plugins.

brainstorming
Auto-invoked

brainstorming

You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or…

@obra@obraInstall Skill
ask-matt

ask-matt

Ask which skill or flow fits your situation. A router over the skills in this repo.

skills12d309k
@mattpocock@mattpocockInstall Skill
accessibility
Auto-invoked

accessibility

Design, implement, and audit accessible UI to WCAG 2.2 Level AA across Web, iOS, and Android — semantic ARIA roles and labels,…

ecc13d293k
@affaan-m@affaan-mInstall Skill
compiler-commit

compiler-commit

Use when you want to verify compiler changes and commit with the correct convention. Runs tests, lint, and format, then commits…

react2d252k
@facebook@facebookInstall Skill

Questions about whiplash.

Which coding agents can use whiplash?
whiplash installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.