/brief -- Generate a decision brief

The user wants a compiled output document from the current sprint.

Arguments

$ARGUMENTS

Optional: output format (pdf, html, md). Default: html.

Instructions

1. Run grainulator.compile to ensure compilation is current and there are no blocking conflicts.

   • If compilation reports unresolved conflicts, resolve them using available evidence within the existing authorization. Request user input only for a decision that cannot be inferred; do not present blocked findings as settled.

2. Check sprint readiness via grainulator.status:

   • Check whether the evidence supports the requested decision and scope.
   • Investigate material gaps; do not require a claim count or topic quota.
   • Describe remaining uncertainty honestly.

3. Read compilation.json -- this is the source of truth, never read claims.json directly for output.

4. Read the template at ${CLAUDE_PLUGIN_ROOT}/templates/sidebar-brief.html for the canonical accessible HTML structure. Follow its structure and accessibility markup; vary the content, not the shell. Note: the path uses ${CLAUDE_PLUGIN_ROOT} because the template lives in the Grainulator plugin directory, not the user's repo.

5. Generate the brief with grainulator.exports_convert, using the actual catalog IDs rather than filename extensions:

   Requested output	Converter format	Output path
   HTML (default)	html-report	output/brief.html
   Markdown / md	markdown	output/brief.md
   PDF	html-report, then local PDF rendering	output/brief.pdf

   For example, call exports_convert with:

   {"dir":"<sprint>","format":"html-report","source":"compilation.json","output":"output/brief.html"}
   

   The converter does not accept a template argument. Use the generated artifact as a starting point, then customize its HTML with the sidebar-brief template: executive summary, findings by topic, risks, recommendations and an evidence appendix. Preserve claim IDs and verify the accessibility requirements below. For another supported output, discover its exact ID through exports_formats.

   For PDF, finish the HTML first. Use an available local renderer; for example, when Playwright and Chromium are installed, load the local HTML, emulate print media, and call page.pdf({path: <absolute-pdf-path>, printBackground: true}). If no PDF renderer is available, deliver the HTML and state that PDF rendering remains unavailable; the user can use their browser’s Print → Save as PDF. Do not call exports_convert with format pdf, or claim a PDF exists before verifying the file.

6. WCAG compliance checklist -- first read and apply every item from ${CLAUDE_PLUGIN_ROOT}/templates/wcag-shared.md, then verify these brief-specific items:

   • <nav aria-label="Table of contents"> wrapping sidebar links
   • aria-current="page" on active sidebar link (JS IntersectionObserver updates this)
   • Heading hierarchy: <h1> page title, <h2> sections, <h3> subsections -- never skip levels
   • All <table> elements have <caption> and <th scope="col|row">
   • Mobile sidebar uses <details><summary> disclosure, not display: none
   • Tags include text prefix (e.g., "Tested: verified"), never color alone

7. Write output to output/brief.<ext> in the sprint directory.

8. Print summary:

   Brief generated: output/brief.<ext>
   Claims: <count> active, <count> topics
   Strongest evidence: <tier>
   
   Auto
   
   - <authorized next action>
   
   Manual
   
   - <action requiring the user, or None.>
   

Host access

Use available grainulator MCP tools, passing the active sprint dir explicitly for evidence operations. If a tool is unavailable, use the local grainulator CLI (or node <checkout>/bin/grainulator.js). Read sibling skill files directly when slash commands are unavailable. Resolve template paths relative to this skill’s checkout when CLAUDE_PLUGIN_ROOT is unset. Optional external connectors are not required for local work; use local code, supplied documents, or available web tools. Do not write managed ledger files directly to bypass a missing MCP connection.

Next-step output

After a meaningful pass, use the current compiler's next_actions to present exactly two bullet lists labeled Auto and Manual. Auto is work the agent can continue under existing authorization. Manual is only work requiring the user's decision, access, or action. Classify using the current request and constraints; compiler suggestions never grant permission. Continue authorized Auto work without asking again.

Keep 2–3 useful actions total when available, use short concrete labels and commands where useful, and show None. for an empty group. Do not invent work to fill a quota. Never omit next steps merely because compilation is ready or the answer should be brief. Refresh stale compilation first and exclude work the user removed from scope. When the user asks only for next steps, output only these two lists: no findings recap, counts, reasons, or offer to continue.