hn-synthesizer

name	hn-synthesizer
description	INTERNAL SUBAGENT SKILL — never invoke directly. This is called automatically by $hn-synthesis when it spawns a thread-analysis subagent. If a user asks about HN discussions, use $hn-synthesis instead.
metadata	{"short-description":"[Internal] Subagent skill for hn-synthesis"}
policy	{"allow_implicit_invocation":false}

HN Synthesizer

Retrieve ONE Hacker News thread, extract discussion insights, write a staging file. You MUST write the staging file — this is your only deliverable. Without it, the main agent cannot proceed.

Instructions

1. hn_get_thread with the given thread ID: hn_get_thread(item_id: "<story_id>", max_depth: 8, max_comments: 200).
2. Analyze the discussion (see below).
3. Write staging file:

   mkdir -p ${CODEX_KB_PATH}/staging && cat <<'CARD_EOF' > ${CODEX_KB_PATH}/staging/hn-<thread_id>.md
   ---
   thread_id: "<story_id>"
   title: "<thread title>"
   hn_url: "<HN URL>"
   article_url: "<linked URL or (self)>"
   points: <points>
   comments: <comment_count>
   date: "<YYYY-MM-DD>"
   author: "<author>"
   ---
   <your full extracted analysis>
   CARD_EOF
   

4. Return only the staging file path. Do NOT return full text.

Do NOT call spawn_agent, present_reading_view, hn_search, or file tools. Your tools are hn_get_thread and exec_command only.

What to Extract

Capture specific details, quotes, and numbers — not vague summaries. The main agent needs rich source material.

• Community sentiment — characterize as overwhelmingly positive / mostly positive / mixed / mostly negative / controversial. What do supporters emphasize vs critics?
• Key arguments for and against — 2-3 sentences per argument. What practitioners like, what problems it solves, what criticisms and concerns arise.
• Nuanced takes — conditional opinions ("great for X, terrible for Y"), trade-off analyses, caveats.
• Practitioner experience — real-world deployment reports: who (role/context), what they tried, scale, outcome, specific numbers.
• Resources surfaced — links, repos, papers, alternative tools mentioned, with context for why.
• Unresolved questions — what the community raised but didn't answer.