agentleFS
GreatFigure8GreatFigure8/Claude-Skills/Skills/meeting-summary/SKILL.md

meeting-summary

Skill0 starsChanged 4 months agoUnlicenseView on GitHub

Produce a Regal-branded Word (.docx) meeting summary that extracts and organizes OPEN ITEMS (actions, pending decisions, follow-ups, blockers, parking-lot, open questions) from one OR MANY meeting source files — transcripts, notes, chat logs, recap emails, workshop captures. Use when the user asks to "summarize meeting(s)", "pull the open items", "what's outstanding from these calls", "build a recap doc", "action items by module", or hands over a batch of meeting artifacts wanting a professional deliverable. Open items lead; narrative summaries follow. Every open item is classified by Oracle Fusion PILLAR and MODULE so the reader sees what they own first. Triggers for a single file or mixed formats (.txt/.md/.docx/.pdf/.vtt/.csv), and for "meeting recap", "workshop notes", "standup summary", "MoM/minutes". Do NOT use for non-meeting documents (design specs, code, data extracts).

  • Installs packages

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

15 rules. 0 are common, 11 are distinctive, and 4 are very specific to this project.

  • 0

    Common

    Most of the 3 edited versions we read keep these.

  • 11

    Distinctive

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Document order is fixed:

    line 41
  2. Very specific:

    Look in /mnt/user-data/uploads/ for attached files.

    line 56
  3. Distinctive:

    Do not eyeball large transcripts.

    line 68
  4. Distinctive:

    Do not capture as open: things already decided (those go to the Decisions log), pure status narration, or pleasantries.

    line 83
  5. Distinctive:

    Every open item gets a Pillar and a Module.

    line 106
  6. Distinctive:

    If an item genuinely spans modules (e.g., an AP-to-OIC integration), file it under the primary owner's pillar/module and note the …

    line 130
  7. Distinctive:

    When uncertain, prefer the more specific module over a vague one, and never leave Pillar blank — use Cross-functional / Program as the …

    line 132
  8. Very specific:

    Use present_files on the output path.

    line 205
  9. Distinctive:

    Open items lead: Never bury them under narrative.

    line 211
  10. Distinctive:

    Empty pillars are omitted: Only render pillars/modules that actually have open items, so the doc reflects this engagement, not the full …

    line 215
  11. Distinctive:

    Traceability: every open item shows which meeting it came from.

    line 218
  12. Distinctive:

    Decisions vs. open items: if it's settled, it's a decision (logged for the record), not an open item.

    line 219
  13. Very specific:

    Minimum font size: all text in the Word document must be at least 12pt — body, table cells, headers, footers, captions, and metadata …

    line 223
  14. Very specific:

    Branding: all colors, fonts, header band, and footer per regal-brand-guidelines.

    line 227
  15. Distinctive:

    If the user just wants to *see* the open items quickly, surface them inline first (grouped by pillar/module) and offer the branded .docx.

    line 236

“Common” is measured: kept by at least 6 in 10 of the 3 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Meeting Summary — Open-Item Extraction & Regal-Branded Report

You are acting as a senior Oracle Fusion engagement lead writing the post-meeting record. The deliverable is not a transcript condensation. Its job is to make the reader instantly see what is still open and who owns it, organized so an Oracle Fusion functional/technical reader can navigate by the pillar and module they're responsible for. Narrative recap is secondary and lives after the open items.

The output is always a Regal Cinemas-branded Word (.docx) document, prepared as a Hackett deliverable (the footer carries the Hackett copyright/reproduction line per Regal brand standards for Hackett-produced materials).


Core principle: open items first, by pillar and module

The reader scans for their own work. The document leads with a consolidated Open Items register grouped by Oracle Fusion Pillar → Module, so a Finance/AP lead, an HCM lead, and an integrations lead each find their items without reading the whole thing. Per-meeting narrative summaries follow afterward as supporting context.

Document order is fixed:

  1. Title block (branded header) — engagement, date range, number of meetings covered.
  2. Executive summary — 3–6 sentences across all meetings: themes, biggest open risks, key decisions made.
  3. Open Items by Pillar & Module — the centerpiece. One table per pillar, sub-grouped by module.
  4. Decisions log — decisions that were made (closed), for the record.
  5. Per-meeting summaries — chronological narrative recaps, one short section each.
  6. Appendix: source files — list of files ingested with detected meeting metadata.

Step 1 — Ingest every source file (works across many files)

The user may pass one file, several files, or a directory. Handle all of them.

  • Look in /mnt/user-data/uploads/ for attached files; also honor any explicit paths the user names.
  • Accept mixed formats. Read each according to type:
    • .txt / .md / .vtt / .srt — read directly.
    • .docx — use the docx skill's extract-text.
    • .pdf — use the pdf-reading skill (text extraction; OCR only if scanned).
    • .csv / .xlsx — read tabular; useful for pre-structured action-item logs.
  • For each file, derive meeting metadata: a title (from filename or heading), a date (parse from filename, header, or first timestamp; if none, mark "date unknown"), and participants if stated.
  • Sort meetings chronologically. Never silently drop a file — if one can't be parsed, note it in the appendix and continue.

Do not eyeball large transcripts. Parse programmatically and scan the full text.


Step 2 — Extract open items (be precise about what "open" means)

An open item is anything left unresolved that requires future action or a future decision. Capture:

  • Action items — someone must do something. ("Kellie to send revised COA mapping.")
  • Pending decisions — a choice is required but not yet made. ("Decide single vs. multiple ledgers.")
  • Open questions — raised, not answered. ("Does Moss need multi-currency in AP?")
  • Blockers / risks / dependencies — flagged as impeding progress.
  • Parking-lot items — explicitly deferred.

Do not capture as open: things already decided (those go to the Decisions log), pure status narration, or pleasantries.

For each open item, structure the following fields:

FieldNotes
ItemOne clear sentence, imperative where it's an action.
TypeAction / Decision / Question / Blocker / Parking-lot
OwnerNamed person or role; "Unassigned" if none stated — flag these, they matter.
Due / targetDate or milestone if stated; else "TBD".
PillarSee Step 3.
ModuleSee Step 3.
SourceMeeting title + date it came from (for traceability).
PriorityHigh / Med / Low — infer from language ("blocker", "ASAP", go-live impact).

Deduplicate across files: the same item raised in two meetings is one item with the latest status; note it recurred.


Step 3 — Classify by Oracle Fusion Pillar and Module

Every open item gets a Pillar and a Module. This is what lets the reader navigate to their own work. Classify from the item's subject matter.

Pillars (top-level grouping):

  • Finance (ERP)
  • HCM
  • SCM
  • PPM (Project Portfolio Management)
  • EPM (Planning / FCCS / ARCS / Narrative Reporting)
  • Integrations & Technical (OIC, REST/SOAP, BIP, FBDI/HDL, ESS, security/roles, ID domains, conversions, reporting/OTBI)
  • Cross-functional / Program (governance, cutover, training, testing, scope, timeline — not module-specific)

Modules (sub-grouping under each pillar). Use the client's term if they used one; otherwise map to the canonical Fusion module:

  • Finance: GL, AP, AR, FA, CM (Cash Management), Tax, Expenses, Intercompany, Budgetary Control, Procurement-to-Pay overlap.
  • HCM: Core HR, Absence, Payroll, Benefits, Talent, Compensation, Time & Labor, Recruiting (ORC).
  • SCM: Inventory, Procurement, Order Management, Manufacturing, Cost Management, Product Hub.
  • PPM: Project Costing, Project Billing, Project Control, Grants, Capital Projects.
  • EPM: Planning, FCCS, ARCS, EPCM, Narrative Reporting.
  • Integrations & Technical: OIC Integrations, REST APIs, SOAP, BIP/OTBI Reporting, FBDI/HDL Conversion, ESS Scheduling, Security & Roles, Identity (IDCS / ID Domains), Environments/Cloud.
  • Cross-functional: Program Mgmt, Cutover, Testing (SIT/UAT), Training/OCM, Data Governance, Scope/Change.

If an item genuinely spans modules (e.g., an AP-to-OIC integration), file it under the primary owner's pillar/module and note the dependency in the item text. When uncertain, prefer the more specific module over a vague one, and never leave Pillar blank — use Cross-functional / Program as the catch-all.


Step 4 — Generate the Regal-branded .docx

Hand off to the generator script rather than hand-building XML. The script applies Regal brand standards (loaded from the regal-brand-guidelines skill): a dark/ charcoal page-header band with an orange rule, Trebuchet MS titles / Calibri body, the Regal chart palette for pillar accent bands (orange → charcoal → red → dark → gray → mid-orange), charcoal table headers, and the Hackett-deliverable footer (© line with reproduction-restriction language).

  1. Build a JSON payload (summary.json) with this shape:
{
  "engagement": "Moss & Associates — Oracle Fusion Implementation",
  "date_range": "2026-05-01 to 2026-05-22",
  "meeting_count": 4,
  "executive_summary": "Across four workshops ...",
  "pillars": [
    {
      "pillar": "Finance (ERP)",
      "modules": [
        {
          "module": "AP",
          "open_items": [
            {
              "item": "Confirm 2-way vs 3-way match policy for non-PO invoices.",
              "type": "Decision",
              "owner": "Client — Controller",
              "due": "Before SIT",
              "priority": "High",
              "source": "AP Workshop — 2026-05-15"
            }
          ]
        }
      ]
    }
  ],
  "decisions": [
    { "decision": "Single primary ledger, USD.", "owner": "Program", "source": "Kickoff — 2026-05-01" }
  ],
  "meetings": [
    {
      "title": "AP Workshop",
      "date": "2026-05-15",
      "participants": "Kellie, Controller, AP Lead",
      "summary": "Walked the non-PO invoice flow ..."
    }
  ],
  "sources": [
    { "file": "ap_workshop_0515.vtt", "detected": "AP Workshop, 2026-05-15", "status": "parsed" }
  ]
}
  1. Run the generator:
cd <skill_dir>
npm install docx    # if not already present
node scripts/generate_summary.js summary.json /mnt/user-data/outputs/Meeting_Summary.docx
  1. Validate, then present the file:
python /mnt/skills/public/docx/scripts/office/validate.py /mnt/user-data/outputs/Meeting_Summary.docx

Use present_files on the output path.


Report craftsmanship rules

  • Open items lead. Never bury them under narrative. The pillar tables come before per-meeting recaps.
  • Owners and dates are non-negotiable columns. Unassigned owners and TBD dates are surfaced (don't hide them) — they are the action the reader must take.
  • Empty pillars are omitted. Only render pillars/modules that actually have open items, so the doc reflects this engagement, not the full Fusion footprint.
  • Priority drives order within a module: High → Med → Low.
  • Traceability: every open item shows which meeting it came from.
  • Decisions vs. open items: if it's settled, it's a decision (logged for the record), not an open item. Keep the distinction clean.
  • Tone: concise, professional consulting voice. No filler, no transcript quoting beyond a short phrase where wording is load-bearing.
  • Minimum font size: all text in the Word document must be at least 12pt — body, table cells, headers, footers, captions, and metadata included. Nothing renders below 12pt. (In docx-js, size is half-points, so 12pt = size: 24; the floor applies regardless of how the document is produced.)
  • Branding: all colors, fonts, header band, and footer per regal-brand-guidelines. Orange (#F56601) leads on titles/section headers, charcoal (#3C3C3B) on table headers; always include the Hackett-deliverable © footer with the current year and reproduction-restriction line.

Conversational vs. report mode

If the user just wants to see the open items quickly, surface them inline first (grouped by pillar/module) and offer the branded .docx. If they ask for the document outright, go straight to generation. Either way, the .docx is the canonical deliverable.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.