before-and-after
Skill1 starsChanged 50 days agoMITView on GitHub
Use when a GitHub PR needs before/after UI stills on the body, or when bulma-qa / the shipper captures those stills with agent-browser.
- Commits and pushes
12 other projects use a version of this.
Adopted by
Most recently taken up by
- NemesLaszlo/claude-code-harnesstheir own version, first seen Nov 8, 2026
- michaelshimeles/rstacktheir own version, first seen Nov 1, 2026
- SwiftyJunnos/skillstheir own version, first seen Oct 20, 2026
- michaelshimeles/skillstheir own version, first seen Oct 15, 2026
- vijpatel7/favorite_skills_and_pluginstheir own version, first seen Oct 15, 2026
- NemesLaszlo/Agent-Collectiontheir own version, first seen Oct 13, 2026
Used with
Projects with it also use greploop-apps, evidence-driven-testing and 3 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
jaydubya818Skillgreploop-apps
Run greploop through @greptile-apps when the normal trigger rejects a large PR. Use only with the companion greploop skill.
Also in 5 projects that use this
jaydubya818Skillevidence-driven-testing
Capture observable runtime proof with video, screenshots, or measured output. Use when a change needs evidence beyond test reports.
Also in 6 projects that use this
BboyfaroukSkillgreploop
Iteratively improves a PR until Greptile gives it a 5/5 confidence score with zero unresolved comments. Triggers Greptile review, fixes all actionable comments, pushes, re-triggers review, and repeats. Use when the user wants to fully optimize a PR against Greptile's code review standards.
Also in 7 projects that use this
DonTiziSkillnew-feature
Plan new features/integrations following SOLID principles. Analyzes codebase architecture, researches official documentation and best practices, designs extensible solution, and creates comprehensive GitHub issue with implementation blueprint.
Also in 6 projects that use this
- +1 more from jaydubya818
Other versions12 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 12
- sanity-io/sanity★ 6.4ksame name
- michaelshimeles/rstack★ 1.4ksame name
- michaelshimeles/skills★ 1.4ksame name
- vercel-labs/before-and-after★ 351same name
- jaydubya818/MissionControl★ 2same name
- mdorf/agent-skills★ 1same name
- Nagacash/nagacodex-factory-skill★ 1same name
- NemesLaszlo/Agent-Collection★ 1same name
- NemesLaszlo/claude-code-harness★ 1same name
- SwiftyJunnos/skills★ 1same name
2 more not listed
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 repository35 other files it gives its agents.
- AGENTS.mdAGENTS.md
- skills/bulma-feature-delivery/adapters/claude.mdCLAUDE.md
- architectSkill
- blast-radiusSkill
- bulma-busSkill
- bulma-code-reviewSkill
- bulma-developerSkill
- bulma-fdSkill
- bulma-feature-deliverySkill
- bulma-goalSkill
- bulma-hostSkill
- bulma-lstmSkill
- bulma-memorySkill
- bulma-qaSkill
- bulma-reviewerSkill
- bulmaSkill
- bulma-triageSkill
- diagnoseSkill
- figure-it-outSkill
- grill-with-docsSkill
- howSkill
- interrogateSkill
- kevinSkill
- no-commentsSkill
- principle-fix-root-causesSkill
- principle-prove-it-worksSkill
- principle-sequence-verifiable-unitsSkill
- receiving-code-reviewSkill
- tddSkill
- technical-writingSkill
- thermo-nuclear-code-quality-reviewSkill
- to-issuesSkill
- to-prdSkill
- typescript-best-practicesSkill
- unslopSkill
The file
We split this skill into the rules it gives an agent
8 rules. 0 are common, 2 are distinctive, and 6 are very specific to this project.
0
Common
Most of the 12 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 15
Run it headless. agent-browser launches its bundled Chrome for Testing process as the rendering engine.
- Distinctive: line 42
Not per-worktree. Base and HEAD share cookies:
- Very specific: line 53
After session restore, open a gated route.
- Very specific: line 81
Stop that app. HEAD worktree, same port, same routes, *-after.png.
- Very specific: line 86
After all captures, agent-browser --session "$SESSION" close.
- Distinctive: line 90
Do not invent a screen.
- Very specific: line 94
Run from the product repo (the cwd gh uses):
- Very specific: line 125
If the body has no block and forge=github, capture the pair (base worktree vs HEAD) and publish as above, then comment.
“Common” is measured: kept by at least 6 in 10 of the 12 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Before and after
Stills on the PR body. Video is QA's job (gh pr comment --attach).
Browser work is agent-browser. This skill owns session, capture
inventory, height alignment, and the marked block.
Run it headless. agent-browser launches its bundled Chrome for Testing
process as the rendering engine; that process is expected. It must include
--headless=new and must not open a visible window or the user's
/Applications/Google Chrome.app. Never use --headed, --auto-connect,
--cdp, --profile, --executable-path, the OS open command, a browser
MCP server, a browser extension, another plugin's browser skill, or computer
use. This rule wins over any skill in the session that says to prefer another
browser. Run every agent-browser command in a shell that ran
ensure-session.sh first: it pins the headless Bulma config.
Load agent-browser skills get core before clicking. Do not copy
vercel-labs/before-and-after (format.mjs is PolyForm Shield). Use
the scripts here (MIT).
When
- Shipper,
forge=github, the diff has UI (routes, screens, widgets, layout/CSS). Aftergh pr create/ update. /qaon a GitHub UI PR whose body lacks the marked block (PR opened elsewhere, or capture failed at ship).
Skip: API-only, --no-pr, forge=git / gitlab, docs/chore with
no screen. Missing agent-browser on a UI PR is BLOCKED, not a
Playwright fallback.
Session
Not per-worktree. Base and HEAD share cookies:
eval "$(../before-and-after/scripts/ensure-session.sh)"
agent-browser --session "$SESSION" --session-name "$SESSION" \
--headed false open "$URL"State: $BULMA_HOME/agent-browser/bulma-<owner>-<repo>/. Never git.
Never Chrome --profile Default.
After session restore, open a gated route. Login form still up → repo helper
(qa:otp / qa:login, docs/ai/qa-login.md, AGENTS.md) → continue
with --session-name "$SESSION" so the next run skips login.
Authenticate until the gated app route loads in that agent-browser session,
then agent-browser --session "$SESSION" --session-name "$SESSION"
--headed false record start with no URL and no --state
(bulma-qa references/EXECUTION.md). record start
opens a new tab. If that tab is Sign in / Check your email,
restore auth with the repo's documented helper into that session and
wait for the app shell before walking. Film and snapshot the same tab. A
login-wall .webm is never PASS.
Inventory
Same surfaces as bulma-qa PLAN.md. One still per changed route (or
each host screen of a shared widget). Desktop always. Mobile only
when the diff touches layout, CSS, media queries, or the design
system. New route: after-only (--before -).
Capture (shipper)
command -v agent-browseror stop (PR still opens; no fake block).eval "$(../before-and-after/scripts/ensure-session.sh)".- Merge-base worktree, start the app (AGENTS.md /
package.jsondev/start). Same origin and port as HEAD, in sequence. - Restore session. Screenshot each surface into
$CAPTURE_DIR(desktop-before.png, …). Paths: no whitespace. - Stop that app. HEAD worktree, same port, same routes,
*-after.png. - Equal height: read
document.documentElement.scrollHeighton both, pad below the shorter page,--fullscreenshot. Confirm pixel height matches. Component diffs: same selector, no padding. - Reject login walls, blank frames, app errors, loading skeletons.
- After all captures,
agent-browser --session "$SESSION" close.--session-namekeeps auth for the next run.
Cannot start the app or auth helper: open the PR without the block. Do not invent a screen.
Publish stills
Run from the product repo (the cwd gh uses):
../before-and-after/scripts/format.sh \
--body-file /tmp/pr-body.md \
--before "$CAPTURE_DIR/desktop-before.png" \
--after "$CAPTURE_DIR/desktop-after.png" \
--label Desktop \
> /tmp/pr-body-next.md
ATTACH=()
while IFS= read -r f; do ATTACH+=(--attach "$f"); done < <(
../before-and-after/scripts/format.sh --attach-list \
--before "$CAPTURE_DIR/desktop-before.png" \
--after "$CAPTURE_DIR/desktop-after.png"
)
gh pr edit "$PR" --body-file /tmp/pr-body-next.md "${ATTACH[@]}"Marker <!-- bulma-before-and-after:start/end -->. Replace that
block only. Place it near the top, after the opening summary, before
Details/Testing. Confirm the body has no leftover ./captures paths.
gh must be ≥ 2.99 (--attach). Older CLI: skip the block, say so.
/qa
Walk the plan with agent-browser. Clip recipe: bulma-qa VIDEO.md. Product errors still FINDINGS → triage → FAIL.
If the body has no block and forge=github, capture the pair (base
worktree vs HEAD) and publish as above, then comment.
Never
- Playwright / Cypress / host MCP as the agent's hands
- Video in the PR body
state saveinside the app repo- Two browser executes at once (QA slot)
- PASS on a UI PR with no video
- PASS on a login-only .webm
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
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.

