resolve

Review and answer ambiguities, or add annotations to enrich cached analysis

Resolve & Annotate

Interactive skill to review annotations and resolve ambiguities recorded during /index. Resolved ambiguities and user annotations enrich cached analysis, improving future /run quality.

Mode

  • Resolve (default): Show unresolved ambiguities for user to answer.
  • Add (--add in $ARGUMENTS): Prompt user to add a new annotation to a specific scope.

Resolve Flow

1. Load Unresolved

Call mcp__ccx__get_annotations(project_dir, annotation_type="ambiguity", unresolved_only=True, limit=20).

If total == 0: Output No unresolved ambiguities. and stop.

2. Present

## Unresolved Ambiguities ({total} total)

1. **{scope}**: {question}
   Context: {content}

2. ...

3. Interactive Resolution

Use AskUserQuestion for each ambiguity one at a time:

  • question: the annotation's question text
  • header: first 12 chars of scope
  • options: 2 suggested answers (infer from context) + "Skip"

If answered (not "Skip"):

  • Call mcp__ccx__resolve_ambiguity(project_dir, scope, question, answer).

Continue until done or user stops.

4. Summary

Resolved: {N} | Remaining: {total - N}

Add Flow

1. Ask Scope

Use AskUserQuestion: "Which scope do you want to annotate?"

  • Options: suggest recent or top-level scopes, or let user type.

2. Ask Type

Use AskUserQuestion: "What type of annotation?"

  • Options:
    • "domain" — Business context, domain knowledge
    • "architecture" — Design rationale, system role
    • "usage" — How to use, gotchas, constraints

3. Ask Content

Use AskUserQuestion: "What's the annotation content?"

  • Let user type freely.

4. Save

Call mcp__ccx__add_annotation(project_dir, scope, annotation_type, content, added_by="user"). Output: Added {type} annotation to {scope}.

Rules

  • Every answer comes from the user — do NOT fabricate answers.
  • Present one ambiguity at a time.
  • Annotations persist in cache across sessions.

Arguments: $ARGUMENTS