Resolve

Finalize a revision by applying the plan from preceding tags to the original content.

Input

• File path: the file being processed.
• Surrounding text: the text before the tag chain, plus the full chain of tags (e.g., <@plan: ...><@output: ...><@resolve> or a multi-turn chain like <@plan: ...><@output: ...><@plan: ...><@output: ...><@resolve>).
• Prompt: typically empty for resolve.

The tag chain may contain a single <@plan>/<@output> pair, or a multi-turn conversation consisting of multiple <@plan>/<@output> pairs. Each pair represents one turn of iterative refinement.

Workflow

1. Read the original content (the text before the tag chain).
2. Read the full conversation history — all <@plan>/<@output> pairs in sequence. The final <@output> represents the most refined plan, but earlier turns provide context for intent and constraints.
3. Apply the plan to the original content. When multiple turns exist, the last <@output> takes precedence, but use earlier turns to resolve ambiguity about what the user intended.
4. Write the result back to the file: replace the entire block (original content + all tags) with the revised text.

Scope Modifiers

The original content may contain scope modifier tags that constrain which regions the plan can be applied to.

• Protect << >>: Text inside << and >> must NOT be modified, even if the plan suggests changes to it. Remove the << >> delimiters in the output, but keep the content unchanged.
• Field (( )): ONLY the text inside (( and )) may be modified by the plan. The surrounding text is context only — do not change it. Remove the (( )) delimiters in the output.

If neither modifier is present, the entire original content is editable (default behavior).

Rules

• Follow the plan faithfully.
• Respect scope modifiers: never edit protected regions, only edit field regions when present.
• All tags (<@plan: ...>, <@output: ...>, <@resolve>) and all scope modifier delimiters (<< >>, (( ))) must be removed; only clean text remains.