From beislid
Generates polished local HTML decks for visualizing evidence proofs, code walkthroughs, UI/CLI demos, reviews, workflows, and sourced documentation. Activated by 'show-me' or similar user requests.
npx claudepluginhub sandsower/beislidThis skill uses the workspace's default tool permissions.
Don't tell me. Show me.
Generates self-contained, branded HTML pages for diagrams, architecture overviews, diff reviews, plans, project recaps, comparisons, and data tables. Proactively replaces complex ASCII tables (4+ rows or 3+ columns) with browser visuals.
Generates self-contained HTML slide decks for product presentations. Reads pm-context.md, extracts brand colors from websites, adds keyboard navigation and print-friendly layouts.
Generates single-file HTML presentations and Obsidian Canvas for big release announcements like new products, migrations, or stakeholder updates for non-technical audiences.
Share bugs, ideas, or general feedback.
Don't tell me. Show me.
show-me turns verification, review, explanation, documentation, demos, or subsystem understanding into a polished local HTML portfolio/deck when chat or terminal output is the wrong medium.
Use this for:
Do not use this for:
verify, ship-it, or other flows in v1Classify the deck before gathering evidence. Use one mode, or mixed when multiple apply. Also choose or infer a presentation: visual-deck for explanatory understanding decks, evidence-deck for proof/demo decks, and report for review/docs/code walkthroughs.
verification — prove claims with command/media evidencereview — inspect work or materials visuallycode-walkthrough — explain diffs, file roles, tests, risks, and rationaleui-demo — show UI behavior through screenshots, videos, GIFs, and browser evidencecli-demo — show command behavior through logs, transcripts, or terminal recordingsdocs — visualize rendered docs, validation output, and documentation structureunderstanding — explain a subsystem/topic with sourced facts and interpretationmixed — combine modes for broad changesStatus values:
PASS — evidence supports the stated claimFAIL — evidence contradicts the claim or a check failedINCOMPLETE — some required evidence is missingNOT SHOWN — a claim was not demonstratedNEEDS CAPTURE — media/evidence tooling was unavailable or capture was deferredEXPLANATORY — understanding/documentation deck, not pass/fail verificationCONFLICTING — sources disagree or evidence points in different directionsLOW_CONFIDENCE — explanation is plausible but source coverage is weakIdentify:
If the user explicitly asked to generate/build/make the deck, proceed after this classification. If the request is ambiguous, propose the mode/outline and ask before generating.
Gather only the context needed for the deck:
Distinguish sourced facts from interpretation. Do not invent rationale; if rationale is inferred, label it as inferred.
Default output root:
${BEISLID_STATE_DIR:-~/.local/state/beislid}/show-me/<repo>/<timestamp>/
Repo-local output is allowed only when explicitly requested or configured:
.beislid/show-me/<timestamp>/
Generated decks are local artifacts and should not be committed in v1.
Use available tools opportunistically:
Ask before destructive, mutating, expensive, long-running, or externally visible actions.
If evidence is unavailable, add a visible NOT SHOWN or NEEDS CAPTURE block. Never imply missing evidence exists.
For understanding decks with EXPLANATORY status, bias toward a visual decision deck by default:
Create a local source/evidence directory. Current HTML rendering assumes online access for CDN presentation libraries (marked, DOMPurify, Mermaid, Highlight.js); source JSON, logs, and copied media remain local, and Markdown source is visible as a fallback if libraries fail to load.
show-me/<repo>/<timestamp>/
index.html
show-me.json
manifest.json
assets/
images/
videos/
gifs/
diagrams/
logs/
commands/
The rendered index.html should be polished and human-reviewable:
Return:
index.htmlshow-me.jsonNEEDS CAPTUREDo not claim the subject passes unless fresh evidence in the deck supports that claim.
show-me.json should use this shape or a compatible superset:
{
"title": "string",
"subtitle": "string",
"mode": "verification|review|code-walkthrough|ui-demo|cli-demo|docs|understanding|mixed",
"status": "PASS|FAIL|INCOMPLETE|NOT SHOWN|NEEDS CAPTURE|EXPLANATORY|CONFLICTING|LOW_CONFIDENCE",
"presentation": "report|visual-deck|evidence-deck",
"summary": "string",
"sections": [
{
"title": "string",
"purpose": "string",
"blocks": []
}
],
"assets": [],
"logs": [],
"provenance": {}
}
Sections are chapters of a portfolio/report. They may look like slides when useful, but long-form sections are preferred for technical walkthroughs.
presentation is optional. Omit it to use the renderer defaults.
Presentation defaults:
understanding + EXPLANATORY → visual-deckverification, cli-demo, ui-demo → evidence-deckdocs, review, code-walkthrough, mixed → report unless explicitly setSupported block types and canonical content keys:
markdown — markdowntable — columns, rowscode — code, optional languagediff — diffdiagram — diagram, optional language: "mermaid" for Mermaid source; copied SVG assets still use media fieldscommand-logimagevideogifcallout — text, optional title/toneverdict — status, textneeds-capturefile-role-tablePi tooling normalizes common aliases like content, body, and table headers, but generated source should prefer canonical keys.
Command evidence must include:
Assets must include:
Understanding decks must include:
Failure decks are valid output. If a check fails, generate the deck with FAIL or INCOMPLETE, include the failure evidence, and add suggested next checks.
Record when available:
Do not dump the full chat transcript by default. Include only the context and rationale needed to understand the deck.
Text logs, JSON, and rendered HTML should receive best-effort secret redaction. Note redactions in provenance.
Screenshots, videos, and GIFs may contain sensitive information. Warn about this unless a real redaction tool was used. Do not claim media is redacted unless it is.
If Pi show_me_* tools or /show-me commands are available, use them to create, update, capture, and render the deck. Prefer show_me_run_command for command evidence so stdout/stderr, exit status, timestamps, cwd, logs, and redactions are captured consistently. Prefer show_me_add_asset for existing screenshots, videos, GIFs, and diagrams so files are copied into the deck with hashes, provenance, captions, and media sensitivity warnings. Use show_me_capture_browser_screenshot for browser screenshots when Playwright is already available; if it is missing or capture fails, use show_me_add_needs_capture or the helper's generated NEEDS_CAPTURE block instead of pretending evidence exists. If the tools are not available, generate the files manually using the artifact contract above.
Future Pi tooling should accelerate this skill, not replace its behavior.