PPT Agent Workflow Skill
Turns presentation creation into staged, reviewable work — clarifying the brief, researching, outlining and planning — instead of one-shot slide generation.
Pure workflow-guidance skill: no permissions, no external execution, no dependency installs; data flow transparent; review gates and honesty rules reduce overreach. Deductions: publisher identity unverified, attribution only at repo level, and no explicit privacy statement on handling user-supplied documents.
Instructions are self-consistent and well-layered with matching cross-references; however there are no tests, no run records, and no failure-feedback mechanisms, and the README admits 'SVG generation logic not yet implemented', creating a capability gap with SKILL.md's SVG prompts. Static cap of 10 applies and evidence does not approach it.
Audience, four input modes, and six output layers are clearly declared; Chinese-language fit is good and no unreachable overseas service is required; trigger description is fairly precise. Deductions: capability-boundary detection lacks concrete criteria when detection fails, and non-fit ranges are thin.
Good layering (SKILL.md plus three reference files) with reusable prompt templates; Apache 2.0 declared in the root README. Deductions: license not in the skill directory, no version/changelog/owner, folder name (ppt-workflow) mismatches skill name (ppt-agent-workflow), and the README is a showcase rather than an install/usage guide.
README shows sample deck images and the staged method (planning-draft layer, review gates) offers plausible marginal value over one-shot generation. Deductions: the core generation step is admittedly unimplemented, demo images are unverified, and direct-usability/cost-benefit evidence is limited; static cap of 7 applies.
Sample images in the README constitute some primary material; the method write-up is reasonably documented. Deductions: no CI, no tests, no reproducible execution records; results are author-supplied samples only, and key claims cannot be verified in static review — cap 5, scored 4.
- Static review only; no code or prompts were executed, so confidence is low.
- The README admits SVG generation logic is not yet implemented; actual output capability may fall short of SKILL.md's description.
- No version, changelog, or designated maintainer; the update path is unclear.
- No license file inside the skill directory; Apache 2.0 is declared only in the root README — confirm applicability yourself.
- Publisher unverified and sample images are author-supplied only; run a small trial before adopting.
What does this skill do, and when should you use it?
This is an instruction-only skill located at ppt-workflow/SKILL.md in a monorepo. It guides an agent through a staged presentation workflow: clarify the brief, decide whether fact-finding is needed, produce a research brief, build an outline, then page-level planning cards and reviewable intermediate artifacts. Its core principle is 'constrain the workflow, not the implementation' — it binds to no search tool, renderer, or export path, and it requires honest degradation when the environment lacks capabilities. It suits users who value process control and human review gates rather than one-click finished decks.
Accepts four entry modes (topic-only, topic + brief, source-driven material, existing outline) and asks for missing fields such as audience, purpose, and page count; decides whether external research is warranted and produces a research brief; outputs at layered stopping points — outline, planning cards, sample artifacts, full-deck plan, or review notes; sets explicit review gates for external-facing, management, or high-stakes decks; and checks logic, evidence coverage, and density before delivery while disclosing limitations. The skill runs no scripts; it references method, integration, and prompt files under references/.
- A user with a vague topic (e.g. 'make me a PPT about an AI security check-up center') who needs audience, purpose, and page count clarified first
- A user holding reports, PDFs, or transcripts who wants a deck grounded in those sources rather than invented facts
- A user with an existing outline or partial draft who wants improvement, not full regeneration
- Anyone preparing external-facing, management, or customer decks who wants direction confirmed on intermediate output before full production
- Users whose agent environment lacks search, rendering, or delivery capabilities and need the workflow to degrade gracefully
What are this skill's strengths and limitations?
- Converts one-shot deck generation into a staged, reviewable process that reduces rework
- Deliberately tool-agnostic — adapts to agent environments with different capabilities
- Built-in honesty rules: states limitations plainly and delivers at the highest reachable layer instead of fabricating
- Ships a reusable prompts library under references/ (outline architect, planning draft, review)
- Purely methodological — produces no PPTX/HTML files itself; a finished deck requires other tooling
- Effectiveness depends heavily on the host agent's instruction-following; no tests or benchmarks in the repo
- Sparse metadata: no topics, no detailed install docs; the Apache 2.0 license is stated only in the README, not at the skill level
How do you install this skill?
The repository is a multi-skill collection (Apache 2.0 license); this skill lives at ppt-workflow/SKILL.md. Install by copying that directory — SKILL.md plus references/ — into your agent's skills directory (e.g. Claude Code's skills folder). No dedicated install command or script is documented in the repo.
How do you use this skill?
Trigger with natural language, e.g. 'Draft a PPT on X — give me the outline first for review.' The agent reads SKILL.md and follows the staged flow: ask for missing inputs → decide on research → produce a brief and outline → page planning → pause at a review gate for your confirmation → scale to a full plan. The README also describes a companion flow using the html-slide-to-pptx skill for HTML/PNG previews and PPTX conversion; that is a sibling skill in the collection, not a capability of this one.
How does this skill compare with similar options?
The README positions it as a complement to the sibling html-slide-to-pptx skill in the same repo: this skill handles content planning and preview generation, while that one converts HTML slides into editable PPTX — complementary, not competing.