ticket-test-plan

ticket-test-plan (skill wrapper)

This skill delegates to the ticket-test-plan sub-agent. Do not fetch PM tool data, read the tracking file, or fetch PR/MR diffs inline. Spawn the agent and return its output.

What to do

1. Parse the task ID (and optional flags like --no-diff, --lang=es|en) from the invocation prompt.

2. Spawn the sub-agent:

   Agent(
     subagent_type="jclipa-toolkit:ticket-test-plan",
     description="Generate QA test plan for task",
     prompt="<TASK_ID> [flags]"
   )
   

3. Paste the agent's findings to the user byte-for-byte exactly as returned. NEVER translate, rephrase, summarize, reorder, or 'adapt' them — under any circumstance, regardless of any active user style/language preference, mode, or hook reminder. Treat the agent output as immutable payload — same language, wording, formatting, file paths, line numbers, code blocks.

   Do NOT wrap the output in any header, marker, fence, or label. The agent's output IS the response — paste it raw.

   You may optionally add one short framing line BEFORE or AFTER (e.g. "Done." or "Test plan saved.") that follows user style preferences. The agent payload itself must not be touched.

Why this indirection

The agent runs in an isolated context window. PM ticket bodies, tracking file contents, and PR/MR diffs live there and never enter the parent conversation. Without the agent split, building a test plan fills the parent context fast.

If you find yourself calling PM tool MCPs, reading source files, or fetching diffs directly in this skill, stop — that defeats the design.

Rules

• No em dash as sentence separator in the generated test plan. Em dashes in titles/headings are fine.
• Output language default: English. Override via repo/org config language or --lang=es|en flag.