Project Wrap-Up

You orchestrate a project wrap-up workflow. Follow these steps in order.

Step 1: Compute the output filename

Determine the project folder name and build the report filename:

basename "$(pwd)"

The report will be named <folder-name>-project-wrapup.md in the project root.

Step 2: Save conversations

Run the save script to extract all conversations as markdown:

bash "${CLAUDE_SKILL_DIR}/scripts/save-all-conversations.sh" "$(pwd)" ".wrap-up-temp"

If the script fails (e.g., no conversations found, jq missing), report the error to the user and stop.

Step 3: Distill conversations

Read the distiller agent definition:

cat "${CLAUDE_SKILL_DIR}/agents/conversation-distiller.md"

Launch a subagent using the Agent tool with:

• prompt: The content of conversation-distiller.md (everything after the frontmatter --- block), followed by these specifics:
  • Input folder: .wrap-up-temp/
  • Output file: <folder-name>-project-wrapup.md (the name computed in step 1, in the project root)
• model: sonnet

Wait for the subagent to complete. If it reports an error, relay it to the user.

Step 4: Clean up

Remove the temporary folder:

rm -rf .wrap-up-temp

Step 5: Report to the user

Tell the user:

• The report filename and location
• How many conversations were processed (from the save script output)
• A brief (2-3 sentence) teaser of what the report covers