tech-intel — an embeddable signal pipeline

collect → score → draft (one LLM call) → lint → publish, with a KnowledgeStore for cross-run memory. The pipeline is source/voice/sink-agnostic: every project-specific piece is injected through an adapter Protocol, so you reuse the orchestration + guardrail and plug in what to watch, how to write, and where to send. It drafts only — it never publishes to a social account on its own.

When to use

"汇总一下 X 上 XX 方向 / 收集这几个方向的动态 / monitor these sources and draft posts / turn this feed into a briefing." Anything shaped like watch sources → draft reviewed, traceable content → push a draft somewhere.

Quickstart (zero-key demo)

SKILL=~/.claude/skills/tech-intel
python3 "$SKILL/run.py" --demo          # file in → linted report out, NO api key