docs: add detailed llm-task examples and common mistakes (fixes #26)#31
docs: add detailed llm-task examples and common mistakes (fixes #26)#31Bandwe wants to merge 1 commit intoopenclaw:mainfrom
Conversation
…law#26) - Add complete Jira + LLM summary workflow example - Document correct openclaw.invoke syntax - Add common mistakes section to avoid confusion - Clarify stdin piping vs env var approaches
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9df918e4d2
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| - id: summarize | ||
| env: | ||
| # Pass tickets via env var to avoid shell escaping issues | ||
| TICKETS_JSON: "$LOBSTER_ARG_TICKETS" |
There was a problem hiding this comment.
The daily-standup example says it summarizes tickets fetched in list-tickets, but summarize sets TICKETS_JSON from LOBSTER_ARG_TICKETS even though the workflow defines only team, project, and limit args. In this form, the summarize step does not receive the previous step’s ticket payload, so copy-pasting this example will not actually summarize fetched Jira tickets.
Useful? React with 👍 / 👎.
| TICKETS_JSON: "$LOBSTER_ARG_TICKETS" | ||
| command: > | ||
| openclaw.invoke --tool llm-task --action json --args-json | ||
| '{"prompt": "Summarize the top 5 most urgent tickets for the daily standup: $TICKETS_JSON"}' |
There was a problem hiding this comment.
The --args-json value is single-quoted, so /bin/sh treats $TICKETS_JSON as a literal string instead of expanding the environment variable. As written, llm-task receives the text $TICKETS_JSON in the prompt rather than the ticket JSON, so the documented workflow cannot produce the intended summary.
Useful? React with 👍 / 👎.
1 participant
Summary
This PR addresses #26 by adding comprehensive documentation for calling OpenClaw tools from Lobster workflows, specifically focusing on llm-task usage.
Changes
Motivation
The issue author struggled with unclear syntax for tool calling in workflows and how to pipe LLM output to next steps.
This PR provides copy-paste-ready examples that work out of the box.