Meeting Minutes to Action Items

Purpose

Meetings produce talk. This skill produces tasks. Paste your meeting transcript, notes, or minutes — raw, structured, or anything in between — and it extracts every decision, action item, and information point into a structured table. Results are grouped by person (default) or by priority, with follow-up email and Slack message drafts so everyone knows exactly what happened and what's next.

This is a text-processing skill. It does not record meetings, access your microphone, connect to meeting platforms, or integrate with external tools. It works exclusively with text you manually paste.

Important limitation: AI extraction is not 100% accurate. Owners, deadlines, and dependencies are inferred from language patterns — always verify assignments, especially for high-stakes or deadline-critical tasks.

⚠️ Confidentiality Warning

Meeting content may contain sensitive business information, personal data, trade secrets, or legally privileged discussions. This skill processes text you provide through the AI model's standard session context — no additional encryption is applied.

• Do NOT paste content from confidential board meetings, legal discussions, HR matters, or any meeting covered by NDA without organizational approval.
• Consider anonymizing names and redacting sensitive details before pasting.
• If your session is shared (Discord, group chat, shared workspace), the meeting content will be visible to all participants.

Use This Skill When

Use this skill when you want to:

• Capture "who does what by when" from a meeting transcript or notes.
• Prepare a follow-up email for client or stakeholder meetings within minutes.
• Compile weekly action items from multiple meetings into one dashboard.
• Catch up on a missed meeting — extract your personal action items instantly.
• Prepare sprint review inputs from planning and sync notes.
• Onboard team members by surfacing decision history from past meeting notes.
• Trace accountability — find who was assigned what and when.
• Generate Slack or chat channel summaries for your team.

Do not use this skill to:

• Record meetings — this skill is text-input only. Use a transcription tool first.
• Create legally binding meeting minutes for compliance or regulatory purposes.
• Replace your project management tool (Jira, Asana, Linear) — this produces copy-paste-ready output, not live tracking.
• Drive automated workflows — it produces text, not tool integrations.

Sample Prompts

Users can start with prompts like:

• "Extract action items and decisions from these meeting notes."
• "Turn this transcript into a task list grouped by person."
• "Draft a follow-up email based on this client call."
• "I missed the standup — what are my action items from these notes?"
• "Here are my notes from 3 meetings this week — consolidate all my tasks."
• "What did we actually decide in this meeting? Just the decisions please."
• "Generate a Slack message summarizing action items for the team."
• "Find all unassigned tasks in these meeting minutes."
• "Who has the most on their plate from these meetings?"
• "Create a status report from this week's meeting decisions."

Best Inputs

To get the most accurate extraction, provide:

• Meeting content — paste the full transcript (with speaker labels like "Alice: ..." preferred), notes, minutes, or chat log. Free-form text works but produces less structured output.
• Meeting context (optional) — meeting type ("Weekly sprint planning"), purpose, or project name. This improves decision classification.
• Participant list (optional) — if speaker labels are absent, providing a list of attendees helps with owner attribution.
• Output preference (optional) — by-person (default), by-priority, email-draft, or slack-draft.
• Multiple meetings (optional) — paste separate meeting blocks with labels for multi-meeting aggregation.

Input Quality & Limitations

Workflow

1. Meeting parsing. Identify participants and roles from speaker labels, explicit introductions, or attendee lists. Detect the meeting type (standup, planning, client call, 1:1, retrospective, etc.) from keywords and structure. Extract date and context if available.

2. Content classification. Classify every meaningful statement into one of three categories:

   • Decisions — conclusions, votes, agreements, scope changes, rejections, strategic choices. Mark with rationale and impact where stated.
   • Action items — tasks, assignments, follow-ups, deliverables. Capture owner, deadline, and any stated priority.
   • Information items — status updates, FYIs, announcements, completed tasks, data shared. No action required; these are for record-keeping.

3. Action item enrichment. For each action item:

   • Owner assignment — extract explicitly named owner; if implicit, mark as "Unassigned" and suggest the most likely owner from context with a confidence note.
   • Deadline extraction — detect explicit dates, relative dates ("by Friday", "next week"), and urgency markers ("ASAP", "urgent", "by EOD").
   • Priority inference — based on deadline proximity, urgency keywords, and stated importance. Use: High (urgent/ASAP/this-week), Medium (next-week/standard), Low (whenever/nice-to-have/no deadline).
   • Dependency detection — identify blocking relationships through language patterns ("after X is done", "once Y completes", "blocked by Z", "prerequisite").

4. Conflict detection. Flag potential issues:

   • Overload — one person with many tasks sharing the same deadline.
   • Missing fields — tasks with no owner, no deadline, or ambiguous scope.
   • Dependency gaps — task B depends on task A, but A has a later deadline than B.
   • Dangling references — tasks referencing people not in the participant list.
   • Duplicate/similar tasks — two action items that appear to be the same work.

5. Output generation. Produce the structured report in the format below, including group-by-person table, decision log, information items, alerts, and a follow-up message draft.

Output Format

Return the structured meeting report in this order.

1. Meeting Summary Header

Report the parsed meeting metadata at the top:

# Meeting Action Items — [Meeting Title]

**Meeting:** [type]  |  **Date:** [extracted or "not provided"]  |  **Attendees:** [list]

---

2. Decisions Made

Table format for decisions:

## 📋 Decisions Made

| # | Decision | Rationale | Impact | Source |
|---|----------|-----------|--------|--------|
| D1 | [decision text] | [why it was made] | [what changes] | [who decided, if known] |

When no decisions are found, state: "No formal decisions identified in these notes."

3. Action Items (Grouped by Person)

Default grouping — one sub-section per person, plus an "Unassigned" section:

## ✅ Action Items

### [Person Name] ([N] tasks)

| # | Task | Priority | Deadline | Dependencies | Confidence |
|---|------|----------|----------|-------------|------------|
| A1 | [task description] | 🔴 High / 🟡 Medium / 🔵 Low | [date or "not stated"] | [dependency or "none"] | High/Medium/Low |

When priority or deadline is uncertain, add a brief note: "this week" — inferred from context, verify.

Priority indicators:

4. Unassigned Tasks

### ⚠️ Unassigned

| # | Task | Suggested Owner | Why | Confidence |
|---|------|----------------|-----|------------|
| U1 | [task description] | [name or "anyone"] | [rationale from context] | Low/Medium/High |

5. Information Items

## ℹ️ Information Items

- [Person] shared: [update summary] — no action required
- [Person] reported: [status update] — completed
- [Announcement or FYI item]

6. Alerts & Warnings

## ⚠️ Alerts

- **[Person] has [N] tasks due in [timeframe]** — verify capacity
- **[Task U1] has no owner** — assign before end of day
- **[Task A2] deadline is [date] but dependency [B1] is due [date]** — dependency chain may need adjustment
- **[Task A3] has no deadline** — set one to avoid indefinite deferral
- **[Task X] and [Task Y] appear to be the same work** — consolidate if confirmed

7. Follow-up Message Draft

Generate one or both message formats:

## 📧 Follow-up Email Draft

**Subject:** [Meeting Title] — Action Items & Decisions

Hi [team / client name],

Thanks for [a productive meeting / the discussion] today. Here's a summary:

**Key Decisions:**
- [Decision 1]
- [Decision 2]
...

**Your Action Items:**
- **[Person]:** [task] (by [deadline])
- **[Person]:** [task] (by [deadline])
...

**Unassigned:**
- [task] — please volunteer or I'll assign by [timeframe]

Full details below. Please update status in [tool name] by [day/date].

Best,
[Your name]

---

## 💬 Slack / Teams Message Draft

@channel **Quick recap from [meeting name]:**

**Decisions:** [one-line summary]

**Actions:**
• @[person] — [task] by [deadline]
• @[person] — [task] by [deadline]

⚠️ [unassigned task] needs an owner

Full notes in thread 🧵

8. Multi-Meeting Merge (When Multiple Inputs Are Provided)

When the user pastes notes from multiple meetings, add:

## 📊 Weekly Dashboard (Combined)

| Person | Total Tasks | High | Medium | Low | Earliest Deadline | Overload Risk |
|--------|-------------|------|--------|-----|-------------------|---------------|
| [name] | [N] | [N] | [N] | [N] | [date] | ✅ / ⚠️ / 🔴 |

## 📋 Combined Decision Log ([N] total)

| # | Decision | Meeting | Date |
|---|----------|---------|------|
| D1 | [decision] | [meeting name] | [date] |

Real Task Examples

Example 1: Sprint Planning → Action Items

Input:

Meeting: Weekly Sprint Planning — May 14, 2026
Attendees: Alice, Bob, Carol, Dave, Eve, Frank

Discussed scope for Sprint 24. Alice will update the API documentation for v2.1 by Friday.
Bob is going to complete the auth module refactor — should be done by May 16 latest.
We decided to postpone Feature X to Sprint 24 because the backend dependency isn't ready.
Carol shared that Q1 customer satisfaction scores are up 12%.
Dave reported the CDN migration is complete — nothing more to do on that.
Alice also needs to review Bob's PR for the auth module, but only after Bob finishes the refactor.
We're adopting Playwright over Cypress for E2E testing. Migration starts next sprint.
Bob will write the Playwright migration plan by May 21.
Eve asked Alice to schedule the Q3 roadmap workshop by May 20.
Frank mentioned the CI pipeline needs updating for the new test framework.

Steps (following the 5-step workflow):

1. Parse: 6 participants identified. Meeting type = sprint planning. Date = May 14, 2026.
2. Classify: 2 decisions (Feature X postponed, Playwright adopted). 5 action items. 2 info items (Carol's CSAT, Dave's CDN).
3. Enrich: Alice → 3 tasks (API docs by May 16, PR review after Bob completes [dependency detected], workshop scheduling by May 20). Bob → 2 tasks (auth refactor by May 16, Playwright plan by May 21). Unassigned → CI pipeline update (suggested owner: Frank or DevOps). Priority: auth refactor and API docs = High (deadline within 2 days). PR review = Medium (depends on Bob). Workshop = Low (May 20, non-urgent). Playwright plan = Medium.
4. Conflicts: Alice has 3 tasks with deadlines close together → overload flag. CI pipeline has no owner → unassigned flag. Bob's Playwright plan depends on decision D2.
5. Generate full structured output.

Expected output (abbreviated):

# Meeting Action Items — Weekly Sprint Planning

**Meeting:** Sprint Planning  |  **Date:** 2026-05-14  |  **Attendees:** Alice, Bob, Carol, Dave, Eve, Frank

---

## 📋 Decisions Made

| # | Decision | Rationale | Impact | Source |
|---|----------|-----------|--------|--------|
| D1 | Postpone Feature X to Sprint 24 | Backend dependency not ready | Scope reduced for current sprint | Team consensus |
| D2 | Adopt Playwright over Cypress for E2E testing | Better cross-browser support | Migration starts next sprint | Team decision |

## ✅ Action Items

### Alice (3 tasks)

| # | Task | Priority | Deadline | Dependencies | Confidence |
|---|------|----------|----------|-------------|------------|
| A1 | Update API documentation for v2.1 | 🔴 High | 2026-05-16 | None | High |
| A2 | Review Bob's PR for auth module | 🟡 Medium | 2026-05-17 | Bob completes refactor (B1) | High |
| A3 | Schedule Q3 roadmap workshop | 🔵 Low | 2026-05-20 | None | High |

### Bob (2 tasks)

| # | Task | Priority | Deadline | Dependencies | Confidence |
|---|------|----------|----------|-------------|------------|
| B1 | Complete auth module refactor | 🔴 High | 2026-05-16 | None | High |
| B2 | Write Playwright migration plan | 🟡 Medium | 2026-05-21 | Decision D2 | High |

## ⚠️ Unassigned

| # | Task | Suggested Owner | Why | Confidence |
|---|------|----------------|-----|------------|
| U1 | Update CI pipeline for new test framework | Frank (or DevOps) | Frank raised it; related to D2 (Playwright) | Medium |

## ℹ️ Information Items

- Carol shared: Q1 customer satisfaction scores up 12% — no action needed
- Dave reported: CDN migration complete — no further action needed

## ⚠️ Alerts

- **Alice has 3 tasks due in 2 days** — verify capacity. PR review (A2) depends on Bob completing B1 first.
- **U1 has no owner** — Frank raised it but wasn't explicitly assigned. Clarify by EOD.
- **B2 deadline (May 21) is 5 days after B1 (May 16)** — dependency chain is reasonable but no buffer. B1 slipping delays B2.

## 📧 Follow-up Email Draft

**Subject:** Sprint Planning 2026-05-14 — Action Items & Decisions

Hi team,

Thanks for a productive sprint planning today. Here's a summary:

**Key Decisions:**
- Feature X postponed to Sprint 24 (backend dependency delay)
- Playwright selected for E2E testing starting next sprint

**Your Action Items:**
- **Alice:** API docs update (May 16), PR review for Bob (after his refactor), roadmap workshop scheduling (May 20)
- **Bob:** Auth module refactor (May 16), Playwright migration plan (May 21)
- **Unassigned:** CI pipeline update — please volunteer or I'll assign by EOD

Please update status in Jira by Wednesday.

Best,
[Your name]

Example 2: Client Call → Follow-up Email

Input:

Client call notes — Acme Corp, May 14 2026
Participants: Our team (me + Sarah), Acme (John - CTO, Lisa - PM)

John is happy with Q1 results. They want a new analytics dashboard by June 1.
Lisa is concerned about data source latency — current refresh is ~4 hours.
We agreed to add a Redis caching layer to reduce latency to near-real-time.
Their IT team (Mark) will provide API access to their data warehouse by May 20.
We need credentials and an API spec from Mark.
Sarah will design the dashboard wireframes and share by May 18 for Acme review.
I'll start building once wireframes are approved and API is available.
Next check-in: June 15.
John also asked about SSO integration — we said we'd research options and come back with a proposal.
Budget was not discussed — we need to follow up separately.

Steps (following the 5-step workflow):

1. Parse: Participants = "me", Sarah (our team); John (CTO), Lisa (PM), Mark (IT) — Acme Corp. Meeting type = client check-in. Date = May 14, 2026.
2. Classify: 3 decisions (add Redis caching, next check-in June 15, SSO research needed). 5 action items. 2 info items (John happy with Q1, Lisa's latency concern).
3. Enrich: Action items → Build dashboard (me, June 1, High, depends on wireframes + API access). Design wireframes (Sarah, May 18, High). Caching implementation (us, implied). API access (Mark/Acme, May 20, High). SSO research proposal (us, no deadline, Medium). Budget follow-up flagged separately — not assigned.
4. Conflicts: Dashboard build deadline (June 1) is only 12 days after API access (May 20) → tight timeline flagged. Two interdependent tasks (wireframes must be approved before build; API must be ready before build). SSO proposal has no deadline.
5. Generate dual output: client-facing email (professional, warm tone, 2-sides table) + internal action list with timeline risk.

Expected output (abbreviated):

## 📧 Follow-up Email Draft (Client-Facing)

**Subject:** Acme Corp — May 14 Check-in: Next Steps & Action Items

Hi John and Lisa,

Thank you for the productive check-in today. Glad to hear Q1 results are tracking well.

Here's a summary of what we agreed and next steps from both sides:

| What | Who | By When |
|------|-----|---------|
| Provide API access to data warehouse | Acme (Mark) | May 20 |
| Share dashboard wireframes for review | Sarah (our team) | May 18 |
| Implement Redis caching layer | Our team | Alongside dashboard build |
| Deliver analytics dashboard v1 | Our team | June 1 |
| SSO integration research & proposal | Our team | TBD — will propose timeline |
| Follow up on budget discussion | Both | This week |

**Critical path note:** The June 1 dashboard deadline depends on API access by May 20. That gives us ~12 days for build and testing — tight but achievable if wireframes are approved promptly.

**Next check-in:** June 15.

Please let us know if May 20 for API access works for Mark.

Best,
[Your name]

Internal action list (second output block):

## ✅ Internal Action Items

### Me

| # | Task | Priority | Deadline | Dependencies | Confidence |
|---|------|----------|----------|-------------|------------|
| 1 | Build analytics dashboard | 🔴 High | 2026-06-01 | Sarah's wireframes approved, Acme API access (May 20) | High |
| 2 | Research SSO options + proposal | 🟡 Medium | Not stated | None | High |

### Sarah

| # | Task | Priority | Deadline | Dependencies | Confidence |
|---|------|----------|----------|-------------|------------|
| S1 | Design dashboard wireframes | 🔴 High | 2026-05-18 | None | High |

### Acme (Mark)

| # | Task | Priority | Deadline | Dependencies | Confidence |
|---|------|----------|----------|-------------|------------|
| M1 | Provide API access + credentials + spec | 🔴 High | 2026-05-20 | None | High |

## ⚠️ Alerts

- **Dashboard deadline (June 1) is only 12 calendar days after API access (May 20)** — tight timeline. Verify scope fits.
- **SSO proposal has no deadline** — set one at next check-in.
- **Budget discussion missed** — schedule a dedicated follow-up.

Example 3: Multi-Meeting Weekly Aggregation

Input:

MEETING 1: Team Standup — May 14

Alice: Working on the payment integration. Should be done by Wednesday.
Bob: Stuck on the caching bug — need help from DevOps. Will ping them today.
Carol: Finished the user profile page. No blockers.

MEETING 2: Product Sync — May 14

Decided to cut the "social sharing" feature from Q3 scope — too much risk, not enough user demand.
Agreed to prioritize performance improvements for the next 2 sprints.
Alice will own the performance audit and come back with findings by May 21.
Dave: The new onboarding flow A/B test is showing 23% improvement in signup conversion.
Bob should update the caching documentation once the bug is fixed.
Eve is taking the lead on accessibility improvements — targeting WCAG 2.1 AA by Q3 end.

MEETING 3: 1:1 with VP Eng — May 14

Discussed team morale — generally good, but engineering team wants clearer roadmap visibility.
Agreed: I'll present a public 6-month tech roadmap at the next all-hands (June 5).
VP wants to see perf improvement metrics by end of Q3.
Alice's promotion case — VP is supportive, will discuss with HR next week.

Steps (following the 5-step workflow):

1. Parse three meetings: Standup (3 participants, 5 items), Product Sync (5 participants, 5 items), VP 1:1 (2 participants + VP, 4 items).
2. Classify across all meetings: 4 decisions, 9 action items (some spanning meetings), 4 info items.
3. Enrich across meetings: Cross-reference Alice's tasks — payment integration (standup, Wed) + performance audit (product sync, May 21). Bob's tasks — caching bug + documentation. Check for overlaps and dependencies.
4. Conflicts: Alice has 3 total tasks if counting the roadmap presentation — high workload. Bob's caching doc depends on fixing the bug — dependency flagged. Performance audit connects to VP's metrics request — related work, not duplicate.
5. Generate merged dashboard, per-person lists, and combined decision log.

Expected output (abbreviated):

## 📊 Weekly Dashboard (Combined — Week of May 14)

| Person | Total Tasks | High | Medium | Low | Earliest Deadline | Overload Risk |
|--------|-------------|------|--------|-----|-------------------|---------------|
| Alice | 3 | 1 | 2 | 0 | May 14 (Wed) | ⚠️ Elevated |
| Bob | 2 | 1 | 1 | 0 | May 14 (today) | ✅ Manageable |
| Me | 1 | 1 | 0 | 0 | Jun 5 | ✅ Manageable |
| Eve | 1 | 0 | 1 | 0 | Q3 end | ✅ Manageable |
| DevOps | 1 | 1 | 0 | 0 | TBD (Bob waiting) | ⚠️ Unconfirmed |

## 📋 Combined Decision Log (4 total)

| # | Decision | Meeting | Date |
|---|----------|---------|------|
| D1 | Cut "social sharing" feature from Q3 scope | Product Sync | May 14 |
| D2 | Prioritize performance improvements for next 2 sprints | Product Sync | May 14 |
| D3 | Present 6-month tech roadmap at all-hands (June 5) | VP 1:1 | May 14 |
| D4 | VP will discuss Alice's promotion with HR next week | VP 1:1 | May 14 |

## ⚠️ Alerts

- **Alice has 3 tasks** (payment integration by Wed, perf audit by May 21, plus promotion discussion context) — workload is elevated. Perf audit may need to be done alongside payment work.
- **Alice's performance audit (Product Sync) connects to VP's Q3 metrics request (1:1)** — consider these as related work, not duplicates. Coordinate scope.
- **Bob blocked on caching bug** — dependent on DevOps response. Action item (ping DevOps) was done today but resolution timeline unclear.
- **VP 1:1 items are sensitive** — promotion discussion and team morale feedback should NOT appear in team-wide communications.

## 💬 Slack Message Draft

@channel **Quick recap from today (May 14):**

**Decisions:** Social sharing cut from Q3, performance is priority for next 2 sprints, 6-month roadmap coming at June all-hands.

**Actions:**
• @alice — payment integration by Wed, perf audit by May 21
• @bob — caching bug fix (blocked on DevOps), docs after fix
• @eve — accessibility improvements (WCAG 2.1 AA by Q3)
• @me — tech roadmap for June 5 all-hands

⚠️ Bob needs DevOps help on caching — can someone jump in today?

Safety Boundaries

• Confidentiality first. Never paste board meeting minutes, HR discussions, legal advice, M&A talks, or any NDA-covered content. The skill processes text in the standard AI session context — treat it like an unencrypted chat.

• This is a text processor, NOT a recording tool. The skill does not access your microphone, meeting platform (Zoom, Teams, Meet, etc.), or any device. You must manually provide text content.

• AI-inferred assignments can be wrong. When owners are not explicitly named, the skill may infer them from context (e.g., "who typically handles that"). These inferences carry lower confidence and must be verified before you act on them.

• Dependency detection is heuristic. The skill detects dependencies through language patterns ("after X", "once Y is done", "blocked by Z"). Complex, implicit, or multi-hop dependencies may be missed. Verify critical dependency chains manually.

• Not a project management tool. This skill converts text to tasks. It does NOT track task status, send reminders, sync with Jira/Asana/Linear, or update anything outside this conversation. Copy the output into your actual PM tool.

• Not a replacement for official minutes. The extracted actions, decisions, and assignments are AI-processed interpretations — they do not constitute legally binding meeting minutes. For compliance, regulatory, or contractual meetings, maintain official records separately.

• Language and format sensitivity. Best with English meeting content. Mixed-language meetings (e.g., English + Chinese discussion) may reduce accuracy. Informal, joke-filled, or highly unstructured meetings will produce lower-quality extraction.

• Hallucination risk. The skill may produce plausible-sounding but incorrect assignments, deadlines, or decisions — especially from ambiguous input. Every output field with confidence below "High" should be treated as a suggestion, not a fact.

• Multi-speaker transcript quality matters. Raw transcripts with speaker labels ("Alice: ...", "Bob: ...") produce significantly better results than undifferentiated walls of text. If your transcript lacks speaker labels, provide a participant list to improve attribution.

• No external tool integration. The skill cannot "create Jira tickets," "send this email," or "post to Slack." It produces formatted text drafts ready for you to copy-paste into those tools.

• Sensitive HR and personnel content. If meeting notes contain performance reviews, compensation discussions, promotion cases, or team morale feedback — these should be redacted before pasting. The skill will process them but cannot guarantee appropriate handling of such sensitive content.

Explicit Refusal / Warning Scenarios

Install-First Success Path

1. Copy this skill directory into ~/.openclaw/skills/meeting-minutes-to-actions/.
2. Restart your OpenClaw session or run the skill reload command.
3. Verify the skill is loaded: check that "Meeting Minutes to Action Items" appears in your available skills.
4. Test with a simple prompt:

   Extract action items from this meeting:
   
   Quick sync: Alice will ship the landing page by Friday. Bob to follow up
   with the vendor about pricing. Decided to use AWS instead of GCP for the
   new project.
   

5. Expected result within 30 seconds: A structured output with 1 decision (AWS over GCP), 2 action items (Alice: ship landing page by Friday; Bob: follow up vendor on pricing), clearly separated into Decisions and Action Items sections with owner/priority/deadline columns.

Success criteria for first install:

• User pastes a few lines of meeting notes and receives a structured action-item table.
• Output clearly separates Decisions from Action Items and Information Items.
• Owners and deadlines are extracted when explicitly stated.
• User can immediately copy-paste the output into a task tracker, email, or chat.

Clean Scan Evidence

Before publishing or sharing this skill, verify it passes basic integrity checks:

# Verify only skill files exist (no credentials, logs, or internal documents)
ls -la /path/to/meeting-minutes-to-actions/
# Expected output: SKILL.md, skill.json, ACCEPTANCE.md, agents/ (3-4 files max)

# Verify no CJK text (English-only requirement)
grep -rP '[\x{4e00}-\x{9fff}\x{3400}-\x{4dbf}]' SKILL.md && echo "FAIL: CJK found" || echo "PASS: no CJK"

# Verify no secrets or credentials
grep -riE '(api.?key|token|secret|password|credential)\s*[:=]\s*[\x27\x22]?[a-zA-Z0-9_-]{8,}' . && echo "FAIL: possible secret" || echo "PASS: no secrets"

# Validate JSON if skill.json exists
python3 -c "import json; json.load(open('skill.json'))" && echo "PASS: valid JSON" || echo "FAIL: invalid JSON"

# Count published files (max 10)
find . -not -path './.git/*' -not -name '.git' -type f | wc -l

Version History