$npx -y skills add mucsbr/ppt-agent-workflow-san --skill ppt-workflowGuide an installing agent through a staged PPT or slide workflow without constraining its implementation choices. Use when the user wants a presentation, deck, storyboard, or slide-ready structure from a topic, article, report, or brief, and success depends on some combination of
| 1 | # PPT Agent Workflow |
| 2 | |
| 3 | ## Overview |
| 4 | |
| 5 | Treat PPT creation as staged work, not one-shot generation. This skill should tell the agent **what must happen and when**, while leaving **how to execute** to the installing agent and its environment. |
| 6 | |
| 7 | Do not prescribe specific search tools, browsers, MCPs, renderers, or export paths unless a particular environment explicitly requires one. |
| 8 | |
| 9 | ## Core Principle |
| 10 | |
| 11 | Constrain the workflow, not the implementation. |
| 12 | |
| 13 | This skill should help the agent decide: |
| 14 | |
| 15 | - when the brief is still unclear |
| 16 | - when fresh fact-finding is necessary |
| 17 | - when to stop and let the user review an intermediate result |
| 18 | - when to continue to a fuller draft |
| 19 | - how to respond honestly when capabilities are limited |
| 20 | |
| 21 | ## Capability Awareness |
| 22 | |
| 23 | Before substantive work, determine at a high level whether the current environment can: |
| 24 | |
| 25 | - gather external facts or at least work from user-supplied materials |
| 26 | - create structured outputs such as notes, outlines, or planning drafts |
| 27 | - create reviewable artifacts |
| 28 | - deliver files or artifacts back to the user |
| 29 | |
| 30 | Only use these findings to decide what layer is achievable. Do not turn this section into a tool-selection policy. |
| 31 | |
| 32 | ## Supported Inputs |
| 33 | |
| 34 | Accept any of these entry modes. |
| 35 | |
| 36 | ### 1. Topic-only |
| 37 | |
| 38 | Example: “帮我做一个 AI 安全体检中心的 PPT” |
| 39 | |
| 40 | If inputs are incomplete, ask for the minimum missing fields: |
| 41 | |
| 42 | - audience |
| 43 | - purpose |
| 44 | - page-count range |
| 45 | - preferred tone or style |
| 46 | - whether fresh fact-finding is needed |
| 47 | - whether the user wants intermediate review before full generation |
| 48 | |
| 49 | ### 2. Topic + brief |
| 50 | |
| 51 | Use when the user already provides audience, purpose, style, page count, and key messages. |
| 52 | |
| 53 | ### 3. Source-driven request |
| 54 | |
| 55 | Use when the user supplies URLs, reports, PDFs, notes, transcripts, or other primary material. |
| 56 | |
| 57 | Treat supplied material as the main context. Add external fact-finding only when it will materially improve the result. |
| 58 | |
| 59 | ### 4. Existing outline or partial draft |
| 60 | |
| 61 | Use when the user already has a structure, early deck, or page ideas and wants improvement rather than full regeneration. |
| 62 | |
| 63 | ## Supported Output Layers |
| 64 | |
| 65 | Stop at the layer the user needs. |
| 66 | |
| 67 | 1. **research-brief** |
| 68 | - facts, evidence, open questions, risks, and source references where available |
| 69 | |
| 70 | 2. **outline** |
| 71 | - deck structure, section logic, page titles, and page goals |
| 72 | |
| 73 | 3. **planning-draft** |
| 74 | - page-by-page planning cards that describe intent, information hierarchy, evidence needs, and layout direction |
| 75 | |
| 76 | 4. **sample-artifact** |
| 77 | - one or more reviewable intermediate outputs in any form the environment can actually produce |
| 78 | |
| 79 | 5. **full-deck-plan** |
| 80 | - approved structure plus planning guidance sufficient for full production |
| 81 | |
| 82 | 6. **review-notes** |
| 83 | - issues in logic, evidence, density, consistency, or emphasis |
| 84 | |
| 85 | ## Default Workflow |
| 86 | |
| 87 | ### 1. Clarify the brief |
| 88 | |
| 89 | Collect only the inputs necessary to avoid avoidable misfires: |
| 90 | |
| 91 | - topic |
| 92 | - audience |
| 93 | - purpose |
| 94 | - page-count range |
| 95 | - style or tone |
| 96 | - must-have sections |
| 97 | - must-avoid claims or directions |
| 98 | - whether the user wants fast rough output or staged confirmation |
| 99 | |
| 100 | ### 2. Decide whether fact-finding is needed |
| 101 | |
| 102 | Perform fact-finding before outlining if: |
| 103 | |
| 104 | - the topic depends on current events, market conditions, product details, statistics, or technical facts |
| 105 | - the user explicitly asks for research-backed content |
| 106 | - the user-provided material is obviously incomplete |
| 107 | |
| 108 | If fresh fact-finding is not necessary, proceed from the supplied context. |
| 109 | |
| 110 | ### 3. Gather or organize context |
| 111 | |
| 112 | Use whatever capabilities the installing agent has to gather or organize supporting material. |
| 113 | |
| 114 | If external fact-finding is not possible, be explicit that the result is source-limited and rely on user-provided material instead of inventing certainty. |
| 115 | |
| 116 | If needed, read `references/agent-integration.md` and `references/prompts.md` before doing this step. |
| 117 | |
| 118 | ### 4. Produce a research brief when research matters |
| 119 | |
| 120 | Before building the deck, summarize the working knowledge into a compact brief: |
| 121 | |
| 122 | - key facts |
| 123 | - supporting evidence |
| 124 | - audience-relevant context |
| 125 | - risks, caveats, and unresolved questions |
| 126 | |
| 127 | This brief can be short, but it should make later outlining more grounded. |
| 128 | |
| 129 | ### 5. Generate the outline before design |
| 130 | |
| 131 | Create the deck structure only after the brief and context are good enough. |
| 132 | |
| 133 | Read `references/prompts.md` and use the “Outline Architect” prompt when appropriate. |
| 134 | |
| 135 | Requirements: |
| 136 | |
| 137 | - preserve logical structure |
| 138 | - keep claims aligned with available |