project-charter-creation

Project Charter Creation

Reads a source Confluence page (e.g. a discovery or requirements doc) and uses the content to fill in structured sections on a target Confluence page (e.g. an implementation or project brief).

CloudId: renoworks.atlassian.net

Workflow

Step 0 — Greet the user

When this skill loads, greet the user with one randomly selected line from the pool below, then immediately follow it with the fixed copy and questions. Do not add anything between the random line and the fixed copy.

Random greeting pool — pick one at random each time:

Lord of the Rings:

1. "One does not simply walk into a project without a proper charter. Let's fix that. 🧙"
2. "You shall not ship... without documented success criteria. Let's get this filled in! ⚔️"
3. "Even the smallest PM can change the course of a project. Let's start with the charter. 🌄"
4. "A wizard's charter is never empty — it is filled in precisely when it needs to be. 🧙‍♂️"
5. "Not all who wander are lost, but your project might be without a proper charter. Let's sort that. 🗺️"

The Big Lebowski: 6. "The charter abides. Let's fill it in. 🎳" 7. "This aggression against undocumented projects will not stand, man. Let's get this charter sorted. 🥃" 8. "Obviously you're not a golfer. But you are a PM, and your charter needs filling in. Let's go. 🎳" 9. "Yeah, well, that's just like... your undocumented project, man. Let's fix it. 😎" 10. "New charter? Far out. Let's see what's in that discovery doc. 🛋️"

Seinfeld: 11. "What's the deal with empty Project Charters? Let's fill this one in. 🧾" 12. "No charter, no project. Pretty, pretty, pretty important. Let's go! 🎤" 13. "Not that there's anything wrong with an empty charter... actually, there is. Let's fix it. ☕" 14. "These pretzels are making me thirsty, and that empty charter is making me anxious. Let's fill it in. 🥨" 15. "It's not a lie if the charter gets filled in. Let's make that happen. 📋"

Star Wars: 16. "Help me fill in this charter — you're my only hope. 🌟" 17. "I find your lack of documented objectives disturbing. Let's sort that out. 🌑" 18. "Do or do not fill in this charter — there is no try. Let's do. ⚡" 19. "These aren't the success criteria you're looking for... actually, they are. Let's write them. 🤖" 20. "The Force is strong with this discovery doc. Let's put it to work. 🚀"

Fixed copy (always follows the greeting):

"{Random greeting} I'll need two things to get started:

Step 1 — Get source and target page URLs

If the user has not already provided both URLs, ask:

1. "What is the URL of the source page (e.g. discovery doc, requirements doc)?"
2. "What is the URL of the target page to fill in?"
3. "Which sections would you like me to fill in, or shall I fill all empty ones I find?"

Extract the numeric page ID from the URL path (e.g. .../pages/3670278205/... → 3670278205).

Step 2 — Read both pages in parallel

Fetch both pages simultaneously using mcp__claude_ai_Atlassian__getConfluencePage:

• Source page: use contentFormat: "markdown" for easy reading
• Target page: use contentFormat: "adf" to get the full ADF structure needed for the update

Step 3 — Identify sections to fill

Inspect the target page ADF for empty or placeholder sections. Common sections include:

• Project Objective — often an info panel with an empty paragraph
• Business Objectives — often a bullet list with one empty list item
• Success Criteria — often a bullet list with one empty list item
• Risk Assessment — often a bullet list with one empty list item

If the user specified particular sections, focus on those. Otherwise fill all empty sections you find.

Step 4 — Draft content for each section

Using the source page content, synthesise appropriate content for each section:

Project Objective A concise 2–4 sentence paragraph describing:

• What the feature/project enables or changes
• The current limitation or gap it addresses
• The approach being taken (at a high level)

Business Objectives 3–5 bullet points covering:

• Who benefits and how (end users, clients, AppBuilders, etc.)
• What new capability or control is being introduced
• Any UX or quality improvements in scope
• What existing behaviour must be preserved (no regression)

Success Criteria 5–8 bullet points that are specific and testable, covering:

• Core functional outcomes (feature works as expected per role/user type)
• Data persistence and API propagation
• UI/UX conditions (labels, flows, states)
• No-regression conditions
• QA/testing pass condition

Risk Assessment 4–6 bullet points, each formatted as: [Risk name]: description of the risk and why it matters. Mitigation: specific action to reduce the risk.

Common risk categories to consider: cost implications, technical complexity/new patterns, UX edge cases, auth/session state, unintended side effects of the change.

Step 5 — Present draft and confirm

Show the drafted content for each section clearly, grouped by section name.

Ask: "Does this look right? Any changes before I update the page?"

Wait for explicit confirmation before proceeding.

Step 6 — Update the target page

Do not manually construct ADF JSON. Use the file-based Python workflow instead:

Pre-flight check — paragraph-in-table

Before modifying the ADF, inspect the fetched body for any paragraph node that appears directly inside a tableBody or table content array (alongside tableRow nodes). Confluence accepts this on read but rejects it on write (400 error). Fix: move such nodes into the nearest layoutColumn's content array instead.

Modification workflow

1. Save the fetched ADF body to /tmp/target_adf.json via json.dump
2. Load it back and make targeted mutations — find nodes by type or attrs.localId, replace their content arrays in place
3. Empty paragraph: replace content with [{"text": "...", "type": "text"}]
4. Empty bullet list: replace the single empty listItem with populated listItem nodes, each with a fresh unique localId (e.g. "sc-02", "risk-03")
5. Risk Assessment items: inline bold for risk name, plain text for description + mitigation — no nested lists
6. Write result to /tmp/adf_body_fixed.json using separators=(',', ':')
7. Validate: node -e "JSON.parse(require('fs').readFileSync('/tmp/adf_body_fixed.json','utf8')); console.log('VALID')"

Preserve all existing localId values for unchanged nodes. Never truncate or omit existing sections (tables, RACI, deliverables, layout).

Call mcp__claude_ai_Atlassian__updateConfluencePage with:

• contentFormat: "adf"
• body: contents of /tmp/adf_body_fixed.json (read via Bash)
• versionMessage: a short description of what was filled in (e.g. "Filled in Project Objective, Business Objectives, Success Criteria, and Risk Assessment")

Step 7 — Confirm and share link

After a successful update, confirm to the user:

• Which sections were filled
• The version number of the updated page
• A direct link to the page: https://renoworks.atlassian.net/wiki/spaces/.../pages/{pageId}/...

Notes

• Always read the target page in ADF format — the full document must be reconstructed and sent back in its entirety; partial updates are not supported
• Never truncate or omit existing sections (tables, RACI, deliverables, team structure) when reconstructing the ADF body
• If the source page mentions cost implications, UX limitations, or technical caveats, these should be surfaced in the Risk Assessment
• If the source page explicitly names an approved implementation option (e.g. "Option 2"), base the Project Objective and approach description on that option
• If the target page has sections not listed above (e.g. "Assumptions", "Out of Scope"), ask the user whether to fill those too