Task Skills Mapper

Reads a task list in any format (.taskmaster/tasks.json, Taskmaster JSON, markdown task list, plain array) and enriches each task and subtask by appending a suggested_skills and suggested_agents block to the details field. The goal is to give the developer or orchestrator immediate clarity on which specialist capabilities are needed before execution begins.

Process

1. Ingest and Parse

Accept the task structure in whatever format is provided:

• .taskmaster/tasks.json (Taskmaster format with tasks[] and nested subtasks[])
• A flat array of task objects
• A markdown list with task descriptions
• A single task description pasted as text

Extract the following fields from each task/subtask to use as analysis input:

• title
• description
• details
• testStrategy (when present)

If the input is JSON, output enriched JSON. If the input is markdown, output enriched markdown. Match the user's format.

2. Analyse Each Task and Subtask

For each task and each nested subtask, reason over the combined text (title + description + details + testStrategy) and apply the mapping tables in references/mapping.md.

Reasoning approach:

• Identify the primary domain (security, testing, refactoring, documentation, architecture, etc.)
• Identify the type of action (implement, review, audit, test, plan, document, research, etc.)
• Match domain + action to the skills and agents in the mapping tables
• Pick only the most relevant ones — don't list everything. Aim for 1–3 skills and 1–2 agents per task. If a task is very broad, up to 4–5 is acceptable, but be selective.

When uncertain between two candidates, prefer the one whose description most closely matches the task's primary deliverable (what must be produced), not just what it touches.

3. Output Format

For JSON input, add two new fields inside each task and subtask. Append to details a clear section separator followed by the suggestions:

{
  "details": "<original details>\n\n---\n**Suggested Skills:** skill-a, skill-b\n**Suggested Agents:** agent-x",
  "suggested_skills": ["skill-a", "skill-b"],
  "suggested_agents": ["agent-x"]
}

Keep suggested_skills and suggested_agents as clean arrays for programmatic use, and the inline append in details for human readability.

For markdown/text input, append to each task entry:

> **Skills:** skill-a, skill-b
> **Agents:** agent-x

4. Preserve Everything Else

Do not modify any field other than details (where you append) and the two new fields. IDs, titles, statuses, dependencies — leave them untouched.

5. Output the Full Structure

Return the complete enriched structure, not a diff or partial. If the input is large (>50 tasks), process in sections and make clear you're doing so.

---

Output Example

Input task:

{
  "title": "Implement Secure Path Normalization for SVG Assets",
  "description": "Refactor local SVG load logic to use filepath.Join and filepath.Abs",
  "details": "Replace string concatenation with filepath.Join() ...",
  "testStrategy": "Test with traversal paths like ../etc/passwd"
}

Output task:

{
  "title": "Implement Secure Path Normalization for SVG Assets",
  "description": "Refactor local SVG load logic to use filepath.Join and filepath.Abs",
  "details": "Replace string concatenation with filepath.Join() ...\n\n---\n**Suggested Skills:** coding-standards, code-review\n**Suggested Agents:** cybersecurity-specialist, developer-specialist",
  "testStrategy": "Test with traversal paths like ../etc/passwd",
  "suggested_skills": ["coding-standards", "code-review"],
  "suggested_agents": ["cybersecurity-specialist", "developer-specialist"]
}

---

Edge Cases

• No details field: Add one containing only the suggestions block.
• Already has suggestions: Check whether they were generated by this skill (look for ---\n**Suggested Skills:**). If yes, replace only that section. If no, append after existing content.
• Very short tasks (title only, no description): Base analysis solely on the title and use the test strategy if available.
• Cross-cutting tasks (e.g., end-to-end tests that touch everything): List the union of all relevant skills/agents but cap at 5 to avoid noise.