Skill
align
Resolve intent, scope, consequential design choices, or important assumptions that need shared understanding before implementation. Use when requested or when…
Development
Skill
/diff
Explain or review a GitHub pull request as an evidence-linked, self-contained offline HTML record. Answer narrow PR questions directly when they do not need a full artifact.
Install
$ npx -y skills add dkstm95/hope --skill diff --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
- Slash command
/diff
Context preview
The summary Claude sees to decide when to auto-load this skill.
Explain or review a GitHub pull request as an evidence-linked, self-contained offline HTML record. Answer narrow PR questions directly when they do not need a full artifact.
SKILL.md
diff.SKILL.mdname: diff
description: Explain or review a GitHub pull request as an evidence-linked, self-contained offline HTML record. Answer narrow PR questions directly when they do not need a full artifact.
Hope Diff
Resolve one exact GitHub pull request, analyze its captured evidence, and report its artifact. Diff covers the captured PR snapshot; local staged, unstaged, and untracked changes are outside its scope.
Resolve the request
Answer a narrow question directly without starting the artifact workflow.
An explicit explanation or review request authorizes the run. Honor existing authorization and display choices without asking again. A bare target without a clear task needs clarification, not an automatic full review.
Use `resolve-target [GitHub PR URL or PR number]` to resolve the target before asking any needed question. Name the repository and PR number. If resolution fails, ask for a URL or number. Once selected, pass that exact target; do not fall back to automatic discovery.
Run adapter commands with `node "<skill-dir>/scripts/cli.mjs"`, replacing `<skill-dir>` with the absolute directory containing this file. In Claude Code it is `${CLAUDE_PLUGIN_ROOT}/skills/diff`. Pass arguments separately; never construct shell commands from PR content.
Analyze the captured evidence
For artifact generation, read `references/workflow.md`, `references/analysis.md`, and `../write/references/writing-standard.md`. Follow the run from `prepare` through `finish` or `cancel`, grounding conclusions in the captured evidence.
Use an independent worker when its perspective is worth the time or the person requests one. Give it a fresh context with the exact request, target, choices, constraints, and Skill path, separate from earlier conclusions. Disclose any limit on requested independence. The same evidence and runtime rules apply whether analysis stays in this conversation or is delegated.
Review generation uses the active host under its data policy; do not claim private PR content stays on the local machine.
Return the artifact
Report the PR, exact head, result scope, absolute HTML path, and any failure or cleanup limit. The artifact completes Diff. The surrounding task handles any publishing, merging, commenting, or code changes under the person's existing authorization.
Maintainers changing the runtime read `references/runtime.md` for its deterministic security and publication contract.
Read more
name: diff description: Explain or review a GitHub pull request as an evidence-linked, self-contained offline HTML record. Answer narrow PR questions directly when they do not need a full artifact.
Hope Diff
Resolve one exact GitHub pull request, analyze its captured evidence, and report its artifact. Diff covers the captured PR snapshot; local staged, unstaged, and untracked changes are outside its scope.
Resolve the request
Answer a narrow question directly without starting the artifact workflow.
An explicit explanation or review request authorizes the run. Honor existing authorization and display choices without asking again. A bare target without a clear task needs clarification, not an automatic full review.
Use `resolve-target [GitHub PR URL or PR number]` to resolve the target before asking any needed question. Name the repository and PR number. If resolution fails, ask for a URL or number. Once selected, pass that exact target; do not fall back to automatic discovery.
Run adapter commands with `node "<skill-dir>/scripts/cli.mjs"`, replacing `<skill-dir>` with the absolute directory containing this file. In Claude Code it is `${CLAUDE_PLUGIN_ROOT}/skills/diff`. Pass arguments separately; never construct shell commands from PR content.
Analyze the captured evidence
For artifact generation, read `references/workflow.md`, `references/analysis.md`, and `../write/references/writing-standard.md`. Follow the run from `prepare` through `finish` or `cancel`, grounding conclusions in the captured evidence.
Use an independent worker when its perspective is worth the time or the person requests one. Give it a fresh context with the exact request, target, choices, constraints, and Skill path, separate from earlier conclusions. Disclose any limit on requested independence. The same evidence and runtime rules apply whether analysis stays in this conversation or is delegated.
Review generation uses the active host under its data policy; do not claim private PR content stays on the local machine.
Return the artifact
Report the PR, exact head, result scope, absolute HTML path, and any failure or cleanup limit. The artifact completes Diff. The surrounding task handles any publishing, merging, commenting, or code changes under the person's existing authorization.
Maintainers changing the runtime read `references/runtime.md` for its deterministic security and publication contract.
Ships withhope
Hope helps people take on more complex work with AI while preserving their judgment and agency.
Get the whole plugin
Stats
8
Stars
1
Forks
Active
Maintenance
JavaScript
Language
MIT
License
8d ago
Last commit
2mo ago
Created
Repo: dkstm95/hope
Other skills on hope.
Skill
design
Design or refine interface mockups with people who describe goals and impressions rather than design terminology. Use for UI direction, visual references,…
Skill
diagram
Create, refine, or review explanatory diagrams and data charts when relationships or quantitative shape communicate better than prose or a small table.…
Skill
sweep
Use only when someone explicitly invokes $hope:sweep in Codex, /hope:sweep in Claude Code, or the host's Hope Sweep command for behavior-preserving codebase…
Skill
toxic-review
Critically review a work product for material flaws and unsupported assumptions, with evidence and proportionate recommendations.
Skill
write
Draft, edit, or review language for clarity while preserving meaning and voice. Apply within implementation and other Skills wherever clearer language helps.