jieli-read

Jieli Read

When to Use

Use this skill immediately when:

• The user pastes or references a Jieli thread URL whose path contains /threads/<thread_id>, including relative links like /threads/T-... and export links like /threads/<thread_id>.md or /threads/<thread_id>.json.
• The user references a raw provider thread id, such as T-....
• The user asks to apply, summarize, continue, compare, or reuse work from a specific known Jieli thread.

Do not use this skill when the user only gives keywords, repo names, file names, or vague clues. Use jieli-find first for search/discovery. Do not use this skill for the current conversation when the needed context is already visible.

Inputs

• A Jieli URL such as https://jieli.app/threads/<provider_thread_id>, https://jieli.app/threads/<provider_thread_id>.md, or a raw provider thread id.
• The user's question or goal for reading the thread. If none is provided, summarize the recent goal, decisions, touched files, commands, verification, and open follow-ups.
• Jieli credentials from JIELI_API_KEY, plugin user config exported as CLAUDE_PLUGIN_OPTION_API_KEY, or ~/.config/jieli/settings.json.
• Jieli uses https://jieli.app by default; no base URL configuration is needed for hosted Jieli.

Environment

Use the jieli-read-thread command for thread reads. It is installed from the plugin bin/ directory and resolves the plugin scripts path itself. If the command is not on PATH, resolve ../../scripts/jieli_helper.mjs relative to this SKILL.md file and run node <resolved-skill-dir>/../../scripts/jieli_helper.mjs read-thread .... Do not call plugin scripts by cache path, enumerate installed helpers, or choose wrapper files in this skill.

If the API key is missing, ask the user to configure the plugin, export JIELI_API_KEY, or write ~/.config/jieli/settings.json.

For hosted Jieli, ask the user to sign in at https://jieli.app, copy an API key, and either:

• Paste the key into the current agent chat and ask the agent to set it.
• Configure the plugin api_key option, which Claude Code exports as CLAUDE_PLUGIN_OPTION_API_KEY.
• Write ~/.config/jieli/settings.json.
• Export it manually:

export JIELI_API_KEY="<api-key-from-jieli.app>"

Read Procedure

1. Resolve the thread id from the URL path or use the raw id. Strip a trailing .md or .json extension from the last path segment.
2. Pass only the provider thread id to the helper script. Do not pass /threads/<thread_id>, a full URL, .md, or .json to the script.
3. Treat a plain page/share URL such as /threads/<thread_id> as a reference to the same thread. The helper reads the agent-friendly markdown export at /threads/<thread_id>.md.
4. Start with a small bounded read for the main content. Do not pull the whole transcript on the first pass:

jieli-read-thread "<thread_id>" --truncate-tool-results --max-chars 12000

5. Only when details are missing, do focused follow-up reads with 1-based inclusive line ranges:

jieli-read-thread "<thread_id>" --truncate-tool-results --start-line 120 --end-line 220 --max-chars 12000

6. Treat the markdown response as the canonical readable thread source. Answer the user's specific question first; do not paste the full transcript unless explicitly requested.
7. Use --truncate-tool-results for the first pass on long threads. This keeps the transcript readable by shortening verbose tool outputs.
8. The helper still applies local --max-chars, --start-line, and --end-line limits.
9. Avoid first-pass reads such as --max-chars 30000 or --max-chars 80000 unless the user explicitly asks for a full transcript. For "main points" or "what happened" requests, prefer the small first read plus line-range follow-ups.
10. If full tool outputs are needed, rerun without --truncate-tool-results, preferably with a focused line range.
11. If structured JSON fields are required, call the helper with --format json and the same local output limits.
12. If the helper fails, report that the thread could not be loaded and include the non-secret error message.

Output

Return a concise context summary with:

• Thread title/id.
• Key user goals.
• Important implementation decisions.
• Files, commands, and verification evidence mentioned.
• Remaining open questions or follow-up tasks.

Do not expose API keys, secrets, or private tokens if they appear in the retrieved transcript.