star-watch-query

name	star-watch-query
description	Answer analytical questions over synced starred repositories using star-watch search/stats, including top-N ranked markdown tables (repo, stars, created date, URL).

Star Watch Query

Use this skill when the user asks questions like:

• "top repos related to X"
• "rank my starred repos by stars/date/topic"
• "show a prettified table from my starred repos"

Defaults

• Run commands from the repository root.
• Prefer markdown table output for ranked/report-style prompts.
• When the user asks for "latest" or "today", refresh first:
  • skills/star-watch-sync/scripts/resync.sh resync

Query Workflow

1) Retrieve semantic candidates

Use semantic retrieval first, then re-rank:

go run ./cmd/star-watch search --json --pool 100 -k 10 \
  --fields "full_name,stars,created_at,url,score" \
  --sort "stars desc" \
  "<topic query>"

Notes:

• --pool 100 means "find top 100 by semantic similarity, then sort by requested field(s)".
• Increase --pool for broader topics.
• Keep score in fields while debugging relevance.

2) Render user-facing table

Preferred table columns for ranking prompts:

• full_name
• stars
• created_at
• url

For direct CLI markdown output:

go run ./cmd/star-watch search --markdown --pool 100 -k 10 \
  --fields "full_name,stars,created_at,url" \
  --sort "stars desc" \
  "<topic query>"

3) Missing created_at handling

If created_at is blank for many rows, run a refresh sync to repopulate metadata:

skills/star-watch-sync/scripts/resync.sh resync