apply-feedback

name: apply-feedback description: "Apply tagged review comments (e.g. [MK], [REV]) in the current document immediately, using per-tag interpretation rules from ~/.mstack/dev/feedback-tags.json. Use when the user says "apply feedback", "apply MK comments", "apply review notes", "apply my edits", or wants tagged comments incorporated directly into a document without an approval gate."

You are reviewing a document that contains feedback comments tagged with one or more configured prefixes (e.g., [MK], [REV]). Tags and their interpretation rules are configured per-user in ~/.mstack/dev/feedback-tags.json.

Step 0: Load feedback tag configuration

Run this bash to check for the config file:

mkdir -p ~/.mstack/dev
CONFIG=~/.mstack/dev/feedback-tags.json
if [ -f "$CONFIG" ] && [ -s "$CONFIG" ]; then
  cat "$CONFIG"
else
  echo "MISSING"
fi

If output is MISSING (first run)

1. Create the file with the default MK entry. Use the Write tool to write ~/.mstack/dev/feedback-tags.json with this content:

   {
     "tags": [
       {
         "tag": "MK",
         "from": "Document author (initials = MK for Mayank)",
         "content": "Direct edit instructions, factual corrections, rephrasing requests, structural changes",
         "action": "Apply edits as written; treat as imperative; only push back if ambiguous"
       }
     ]
   }
   

2. Tell the user verbatim:

   Saved your feedback tag config to ~/.mstack/dev/feedback-tags.json. Pre-seeded with MK as the default. Edit this file directly to add, modify, or delete tags later.

3. Use AskUserQuestion: "Want to add another tag now? You can also add more later by editing the file." with options:

   • A) No, continue with MK
   • B) Yes, add another tag

4. If the user picks B, ask 4 questions in sequence (one AskUserQuestion call each, free-text answers):

   • "What tag prefix? (short uppercase letters only — e.g., REV, FB, TODO. Will be matched case-insensitively in documents.)"
   • "Who are these comments from? (e.g., 'External reviewer')"
   • "What do these comments typically contain? (e.g., 'Suggestions and open questions')"
   • "What action should I take with these comments? (e.g., 'Treat as suggestions; flag for discussion before applying')"

   Read the current config, append the new tag object to the tags array, write the updated file. Then ask "Add another?" again — loop until user says no.

5. Proceed to Step 1 once setup completes.

If output is valid JSON

Parse it. Build a list of all configured tags from the tags array. Each tag has fields tag, from, content, action.

If JSON is malformed, warn the user (e.g., "feedback-tags.json is invalid JSON, falling back to default MK tag for this run"), then use the default MK tag with the same defaults as the first-run pre-seed.

Step 1: Locate all configured tag comments in the document

For each tag from the loaded config, search the document case-insensitively for the bracketed pattern. For example, if tags MK and REV are configured, search for [MK], [mk], [Mk], [REV], [rev], [Rev], etc.

Use the Grep tool with the -i flag, or grep -i "\[mk\]" style commands. Aggregate matches across all configured tags.

If no matches found for any tag, tell the user: "No comments found for any configured tag: TAG1, TAG2, …" listing the configured tag prefixes, and stop.

Step 2: Apply each comment immediately

For each match, apply the requested change without asking permission, using the per-tag action and content fields to inform interpretation:

• A tag whose action says "Apply edits as written; treat as imperative" → just apply the edit.
• A tag whose action says "Treat as suggestions; flag for discussion" → in this apply-feedback skill, still apply (this skill's posture is apply-now), but be more conservative; if the suggestion is open-ended ("consider rewording"), make the most reasonable edit and note it in your final summary.

Apply edits by:

• Updating content based on the comment
• Removing contradicting information
• Adding missing information
• Restructuring sections if the comment requests it
• Removing the bracketed tag (and its inline note) after incorporating the feedback

Step 3: Preserve flow

Ensure changes integrate smoothly with existing content. Preserve overall structure and tone. Match the document's voice when adding content.

Important Rules

• Remove ALL configured-tag brackets after incorporating feedback
• Do NOT ask for permission — apply changes directly (this is the apply-immediately skill)
• Be comprehensive — address every comment for every configured tag
• Maintain consistency throughout the document
• Use each tag's action and content fields as guidance for how aggressively/literally to apply

Example

With MK configured (action: "Apply edits as written"):

Before:

The system will use the brand_mentions table [MK] Rename to response_metrics.

After:

The system will use the response_metrics table

Apply all configured-tag feedback now and update the document.