What does this skill do, and when should you use it?
This skill provides a comprehensive quality-control workflow for investment banking presentations across four dimensions: number consistency (checking metrics across slides and calculations), data-narrative alignment (verifying claims are supported by data), language polish (removing casual phrasing and applying IB terminology), and formatting audit (charts, typography, numbers, dates, footnotes). It automates extraction via command-line scripts and outputs findings categorized by severity to help users elevate professionalism and avoid errors.
Extracts presentation text using markitdown; runs extract_numbers.py to pull numbers with slide references; checks key metric consistency, calculation correctness, unit format consistency, and time-period alignment; maps chart directions and claims to supporting data, verifying market share plausibility; scans for casual language, vague quantifiers, contractions, and exclamation points, offering IB terminology replacements; audits each slide for chart sourcing, axis labels, legends, font hierarchy, number and date formatting, and footnotes; categorizes all findings by severity (critical, important, minor) and produces a report using a reference template.
- An IB analyst performs a final pre-submission check on a client deck, catching number mismatches and formatting issues.
- An investment manager reviews a junior's M&A or financing presentation to ensure data supports the narrative.
- An IB team polishes language in a deck before final delivery, replacing casual phrases with professional terminology.
- A consultant uses the skill for a quick formatting consistency check before a client meeting.
- A financial analyst validates a market share claim against revenue figures to ensure plausibility.
How do you install this skill?
- The skill depends on external tools (markitdown, pptx skill) without installation or security instructions; ensure these dependencies are trusted and available.
- The script extract_numbers.py may have path errors and is untested for abnormal inputs (e.g., corrupt files); verify robustness before use.
- The skill is English-only and may not support Chinese materials; external services may be unreachable in mainland China.
- Static review only; not executed. Use with caution and test in a controlled environment before production use.
- Shell / CLI
Python
The skill is part of the LangAlpha monorepo (ginlix-ai/LangAlpha), which bundles 33 skills. Clone the repository or copy the skills/check-deck/ directory into your workspace's skills folder (following the Agent Skills spec).
tmp="$(mktemp -d)"
git clone --depth 1 https://github.com/ginlix-ai/LangAlpha.git "$tmp"
mkdir -p ~/.claude/skills
cp -R "$tmp/skills/check-deck" ~/.claude/skills/
rm -rf "$tmp"Generated from the source repository and skill path; it copies only this skill's folder. If the author's install steps above differ, follow those first. To scope it to one project, replace ~/.claude/skills with that project's .claude/skills.
How do you use this skill?
Convert the presentation to Markdown before checking: run python -m markitdown presentation.pptx > content.md, then python scripts/extract_numbers.py content.md --check for number consistency. For visual inspection, use the pptx skill workflow to convert slides to images. Finally, generate a report using the provided template in references/report-format.md.
What are this skill's strengths and limitations?
- Covers four critical quality dimensions comprehensively.
- Automates number extraction and consistency checks, saving manual effort.
- Outputs findings categorized by severity for easy prioritization.
- Derived from anthropics/financial-services-plugins, industry-validated.
- Only provides extraction with slide references; no automated fix mechanism mentioned.
- Visual inspection depends on another skill (pptx), requiring integration.
- No built-in test suite; QA relies on user diligence.
- SKILL.md does not detail all script behaviors or failure modes.
How does this skill compare with similar options?
Side by side with related skills; every score comes from the same FSRS standard.
| Skill | FS score | Stars | Last updated | License |
|---|---|---|---|---|
| IB Deck QC this page | 35 · Not recommended | ★ 1.8k | 3d ago | Apache-2.0 |
| Investment Banking Deck Checker ✓ Anthropic · Official | 47 · Use with care | ★ 39k | 20d ago | Apache-2.0 |
| Model Checker for Financial Audits | 53 · Use with care | ★ 1.8k | 3d ago | Apache-2.0 |
| Investment Banking Pitch Deck Filler ✓ Anthropic · Official | 53 · Use with care | ★ 39k | 20d ago | Apache-2.0 |
| M&A Buyer Universe Builder ✓ Anthropic · Official | 52 · Use with care | ★ 39k | 20d ago | Apache-2.0 |
How did FollowSkills review this skill?
The skill is purely documentation and scripts with no network calls or external effects; it relies on third-party tools (markitdown, pptx skill) without security assessment. The script auto-extracts and flags inconsistencies but lacks user confirmation and rollback. No sensitive data is handled, but external dependencies are undisclosed in detail. Deductions for least privilege, confirmation, and verifiable source attribution.
Instructions and script are self-consistent, but depend on external tools and possibly nonexistent script path. Static review cannot verify key paths, and abnormal input handling (e.g., corrupt files) is insufficient. Max static score 10, but lacking tests and error detail reduces to 5.
Skill clearly targets investment banking deck QC; scenario is clear. But trigger conditions (automatic vs manual) are vague, boundaries (e.g., non-English content) not declared. Environment fit: English-only docs, and external dependencies may be unreachable in mainland China, but skill is offline-capable. Score 7.
Structure is clear with external references and license declaration, but lacks versioning, changelog, installation details, examples, and FAQs. Ownership and update path not defined. Score 8.
Provides full QC workflow and output template, but core value depends on markitdown conversion and script accuracy, unverified statically. Manual QC might be more accurate; marginal value not fully proven. Max 7, but no execution evidence, so 5.
Claims scripts and references but no test suite or CI integration. Static review cannot reproduce; no third-party corroboration. Score 3.
Open a dimension to read why it scored that way
Evidence confidence：Low — Mostly static review, author material or a limited demo; useful for discovery, not high-risk decisions.
See the full review method →