Collaboration and Comments
Power Prompt fosters collaborative prompt engineering and continuous quality improvement through contextual commenting and threaded review workflows.
1. Adding a Comment to a Prompt
To initiate an exchange or document qualitative feedback on a prompt:
- Open the prompt in the editor via Mes prompts.
- Locate the Comments block in the right sidebar.
- Type your observation or suggestion in the text area at the bottom of the panel.
- Click the send button (paper airplane icon).
Version Filtering and Contextual Anchoring
To ensure feedback remains clear and relevant as prompts evolve:
- Version Filter: The dropdown selector at the top of the block (e.g.
Current Version ▾) lets you filter comments to display only those linked to the active revision, or view discussions across all versions. - Version Badges: Each comment displays a distinct tag showing the exact version it was posted against (e.g.
Draft of V3), ensuring that notes retain their context even after multiple iterations.
Common Comment Use Cases:
- Proposing an Enhancement: Suggest a more concise formulation or a stricter output constraint.
- Flagging Inaccuracies: Notify the author of unexpected hallucination or behavioural edge cases.
- Requesting Review: Invite peer or legal review prior to production certification.
- Documenting Engineering Rationale: Record why specific few-shot examples or formatting delimiters were chosen.
2. Threaded Discussions and Peer Mentions
To keep discussions structured without cluttering the main prompt instruction canvas:
- Threaded Replies: Reply directly to an existing comment to maintain focused discussion threads.
- Peer Mentions (
@username): Type@followed by a teammate’s name to alert them directly. The tagged collaborator receives an instant notification via the platform interface or email. - Discussion Lifecycle: Each version of a prompt has its own dedicated discussion thread. When a new version is published, the associated thread is reset and starts empty, while exchanges from prior versions remain archived.
This ensures full provenance for prompting decisions across the lifespan of your enterprise library.