• Pain Point Driven — Only collect real pain points and clear ideas, no hypothetical planning
• AI Organizes, Users Decide — AI is responsible for following up, categorizing, and merging; users are responsible for confirmation and final decisions
• Archive First, Decide Later — New ideas go into the backlog first, no rush to set priorities or start development
• Keep the Backlog Active — Regularly clean up outdated entries, promote entries that have been clarified, and prevent the backlog from becoming a garbage dump
• Stay Within Boundaries — Backlog management stops at "ready to implement"; subsequent design/PRD/development is handled by other processes

Jot down anytime, add entries whenever, review periodically.

Status Description:

• Ready to Implement — Clear pain point, clear understanding of what to build, can enter design/PRD at any time
• Needs Further Discussion — Direction is set, details are not fleshed out, requires research or discussion
• On Hold — Just keep a record, no effort needed for now
• Completed — Done, archive with version number

• Pain Point: xxx
• Solution: xxx
• Notes: xxx

• Direction: xxx
• Dependencies: xxx
• To Be Fleshed Out: xxx

• Idea: xxx
• Notes: xxx

• Completion Date: yyyy-mm-dd
• Brief Description: xxx

undefined

• What is the pain point? — What problem are you currently facing? In what scenario is it frustrating?
• How frequent is it? — How often do you encounter this?
• Current Workaround — If this feature is not implemented, how do you solve it manually?

• Directly writing the user's one-sentence input into the backlog without first following up and confirming
• If the follow-up exceeds 3 rounds without convergence, archive it based on current understanding first, marked as "Needs further discussion"

1. Read Existing Backlog (if it exists)
2. Check for Merge Possibility — Does the new idea fall into the same category as an existing entry?
   • Mergeable → Inform the user "This is related to #X, I suggest merging into xxx" and wait for user confirmation
   • Not Mergeable → Treat as a new entry
3. Determine Status — Archive based on follow-up results:
   • User can clearly explain the pain point and general solution → Ready to Implement
   • User has a direction but details are vague → Needs Further Discussion
   • User says "Just keep a record" → On Hold
4. Confirm Categorization Result with User, then write to the file after user approval

• AI determines the status without confirming with the user
• Forcibly merging requirements that the user considers different

docs/backlog.md

1. Add entry details in the corresponding section
2. Update the overview table (add a new row or modify an existing one)
3. If a merge occurs, update the content of the merged entry and the overview table

• The overview table and detail section must be updated synchronously; do not modify only one part
• Numbers are incremented sequentially; do not reuse deleted numbers
• Completed entries remain in the overview table with status marked as "Completed"

1. Read the complete backlog file
2. Review each entry and provide suggestions:
   • Outdated → Suggest deletion, explain the reason
   • Clarified → Suggest promotion (On Hold → Needs Further Discussion, or Needs Further Discussion → Ready to Implement), explain the basis
   • Still Vague → Ask 1-2 key questions to help the user clarify
   • No Changes → Skip, no unnecessary comments
3. After the user confirms each change suggestion, update the file in batches

• Batch modify status without user confirmation
• Ask "Do you want to change this?" for every entry (directly skip entries with no changes)

1. Read the backlog and list all entries marked "Ready to Implement"
2. If there are no "Ready to Implement" entries, pick the closest ones from "Needs Further Discussion" and help the user clarify them
3. For candidate entries, analyze using three criteria:

1. Provide analysis results, do not make decisions for the user, let the user choose
2. After the user selects, mark the entry as "In Progress" in the backlog (or the user directly triggers the design-exploration / PRD process)

• Making decisions for the user on which requirement to implement
• Forcibly assigning priority numbers to all entries (only conduct comparative analysis when the user needs to select)
• Starting design or writing PRD immediately after the user selects (the backlog process ends at "selection"; subsequent processes are triggered by the user)

1. Move the entry from the current section to "Completed"
2. Add version number and completion date
3. Update the status in the overview table
4. If the entry is depended on by other entries, check if the dependent entries can be promoted and remind the user

• Make decisions for the user on which requirement to implement
• Start design or writing PRD immediately after the user throws an idea (archive first, start only when the user says so)
• Fabricate requirement content (must be organized based on the user's original words)
• Modify the backlog file without user confirmation
• Forcibly set priorities for all entries (only conduct comparative analysis when the user needs to select)
• Treat the backlog as a to-do list to urge the user (the backlog is passive; only respond when the user takes the initiative)