From you-read-i-read
Use when starting an interactive reading session over a paper. Triggers include "/read", "/read <id>", "/read <keywords>", "let's read something", "I want to read about <topic>", "let's read X", "show me a paper to read". The agent picks from the to-read queue, loads the structured summary, walks through it menu-driven, drills into raw text on request.
npx claudepluginhub zhangzhuosjtu/youreadiread --plugin you-read-i-readThis skill uses the workspace's default tool permissions.
Interactive reading skill. Owns `/read [<id> | <keywords>]`. Loads the
Guides Next.js Cache Components and Partial Prerendering (PPR): 'use cache' directives, cacheLife(), cacheTag(), revalidateTag() for caching, invalidation, static/dynamic optimization. Auto-activates on cacheComponents: true.
Guides building MCP servers enabling LLMs to interact with external services via tools. Covers best practices, TypeScript/Node (MCP SDK), Python (FastMCP).
Share bugs, ideas, or general feedback.
Interactive reading skill. Owns /read [<id> | <keywords>]. Loads the
structured summary, hooks the user with motivation + insight, lets them
choose how deep to go, grounds custom questions in the raw text. Marks
status correctly only when the user explicitly says they're done.
~/.you-read-i-read/data/papers/summaries/<id>.md. If absent, run a deep
summary first via the paper-summarizer subagent and persist before
proceeding.reading → read silently. Status flips only via:
[5] Done — mark read menu choice, ORreading.~/.you-read-i-read/data/papers/raw/<id>.txt (or .html / .pdf) and
quote with (§4.2, Table 3)-style references. If the answer isn't in
the source, say so plainly.signal-log --event read_started on entry,
--event read_finished on confirmed read.Pick the paper.
/read <id> → that paper./read <keywords> → search via
manage_data.py paper-list --status to-read --query "<keywords>",
present top matches, let user choose./read (no args) → recommend the top of the ranked queue from
manage_data.py rank --status to-read --limit 5. Briefly justify
("highest accept_likelihood + trending"). Confirm.Mark reading and log read_started:
manage_data.py paper-update <id> --status reading
manage_data.py signal-log --event read_started --paper-id <id>
Hook. One paragraph synthesizing the summary's Motivation +
Insight. End with the source URL.
Menu. Offer:
[1] Tech details (the method)
[2] Evaluation (numbers + ablations)
[3] Open the paper in browser → <source_url>
[4] Ask my own question
[5] Done — mark read
[q] Pause for now (status stays `reading`)
Drill. On [1] / [2], render the relevant summary section, then
offer to go deeper from raw text. On [4], ground in raw text and
answer with section/page references.
Finish. When the user picks [5] (or signals they're done):
manage_data.py paper-update <id> --status read
manage_data.py signal-log --event read_finished --paper-id <id> --field duration_seconds=<approx>
Optionally accept a free-form note for read_log.
End-of-session prompt. If the user picks [q] or otherwise exits
without [5], explicitly ask before closing:
Did you finish reading <title>? [y]es / [n]ot yet / [a]rchive without finishing.
y → mark read, log read_finished.n → status stays reading, no signal logged.a → mark archived, log archived_without_reading (optional reason).When /read has no argument, justify the pick in one line:
"highest accept_likelihood (0.82); trending +0.45 on HN today" or
"added 6 days ago — going stale". The user can override with
/read <other-id>.
read because the user said "thanks" — wait for explicit
yes / [5].reading forever.