Skip to content

adding SKILL files#14

Closed
kadolor wants to merge 3 commits intomainfrom
adding-skills
Closed

adding SKILL files#14
kadolor wants to merge 3 commits intomainfrom
adding-skills

Conversation

@kadolor
Copy link
Copy Markdown

@kadolor kadolor commented Apr 6, 2026

No description provided.

@kadolor kadolor self-assigned this Apr 6, 2026
Adds the using-wherobots skill that guides agents on when to use CLI,
MCP, DB-API, JDBC, or SDKs. Updates exploring-api and submitting-jobs
with cross-references to the routing skill.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
@notion-workspace
Copy link
Copy Markdown

@notion-workspace
Copy link
Copy Markdown

@notion-workspace
Copy link
Copy Markdown

Create SKILL.md for CLI

Copy link
Copy Markdown

Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull request overview

Adds a set of Wherobots CLI “skill” documents (duplicated under both skills/ and .claude/skills/) plus a Claude plugin manifest, to guide users on submitting, monitoring, logging, and API exploration workflows.

Changes:

  • Added five new skill guides for Wherobots CLI workflows (submit jobs, monitor jobs, view logs, explore API, tool selection).
  • Mirrored the same skill guides under .claude/skills/ for Claude integration.
  • Added .claude-plugin/plugin.json manifest metadata.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 5 comments.

Show a summary per file
File Description
skills/viewing-logs/SKILL.md Adds guidance for fetching/following job-runs logs.
skills/using-wherobots/SKILL.md Adds decision guidance for picking CLI/MCP/DB-API/JDBC/Airflow tools.
skills/submitting-jobs/SKILL.md Adds documentation for job-runs create, storage resolution, and common flags.
skills/monitoring-jobs/SKILL.md Adds documentation for listing runs, status shortcuts, and metrics.
skills/exploring-api/SKILL.md Adds documentation for using wherobots api discovery and common endpoints.
.claude/skills/viewing-logs/SKILL.md Duplicates viewing-logs skill for Claude skills directory.
.claude/skills/using-wherobots/SKILL.md Duplicates using-wherobots skill for Claude skills directory.
.claude/skills/submitting-jobs/SKILL.md Duplicates submitting-jobs skill for Claude skills directory.
.claude/skills/monitoring-jobs/SKILL.md Duplicates monitoring-jobs skill for Claude skills directory.
.claude/skills/exploring-api/SKILL.md Duplicates exploring-api skill for Claude skills directory.
.claude-plugin/plugin.json Adds Claude plugin manifest metadata for the skill set.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment on lines +68 to +76
| Task | Tool |
|---|---|
| "What tables are available?" | MCP (`list_tables_tool`) |
| "Run this spatial query" (interactive) | MCP (`execute_query_tool`) |
| "Run this query in my Python app" | DB-API or JDBC |
| "Submit this script as a batch job" | CLI `job-runs create` |
| "Check my running jobs" | CLI `job-runs list` |
| "Start/stop a SQL session" | CLI `api sql ...` |
| "Manage files in S3" | CLI `api files ...` |
Copy link

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The markdown table in this section is malformed (extra leading | on each row: || ...). This will render incorrectly in most Markdown viewers; use a standard 2-column table format with single leading/trailing pipes (e.g., | Task | Tool | / | --- | --- |).

Copilot uses AI. Check for mistakes.
Comment thread skills/submitting-jobs/SKILL.md Outdated

## Storage resolution order

The script must reference an S3 location. Resolve where it lives or will be uploaded in this order:
Copy link

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This section says the script "must reference an S3 location", but later the command reference explicitly allows a local .py file path and describes auto-upload. Consider rewording to clarify that runs execute from S3, and the CLI will upload local scripts automatically (unless --no-upload is used with an s3:// URI).

Suggested change
The script must reference an S3 location. Resolve where it lives or will be uploaded in this order:
Job runs execute from an S3 location. The CLI accepts either a local `.py` file path or an `s3://` URI; if you pass a local file, it is uploaded automatically to the resolved S3 destination. Resolve where the script already lives or will be uploaded in this order:

Copilot uses AI. Check for mistakes.
Comment on lines +1 to +4
---
name: viewing-logs
description: Fetches or streams logs for a Wherobots job run. Supports following in real time and tailing the last N lines. Use when a user wants to see logs, debug a failed job, tail output, or follow a running job's progress.
---
Copy link

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

These skill docs appear to be duplicated under both skills/... and .claude/skills/.... Keeping two copies in sync is error-prone; consider using a single source of truth (e.g., symlinks or a generation/copy step) so future edits can't drift between the two trees.

Copilot uses AI. Check for mistakes.
Comment on lines +68 to +76
| Task | Tool |
|---|---|
| "What tables are available?" | MCP (`list_tables_tool`) |
| "Run this spatial query" (interactive) | MCP (`execute_query_tool`) |
| "Run this query in my Python app" | DB-API or JDBC |
| "Submit this script as a batch job" | CLI `job-runs create` |
| "Check my running jobs" | CLI `job-runs list` |
| "Start/stop a SQL session" | CLI `api sql ...` |
| "Manage files in S3" | CLI `api files ...` |
Copy link

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The markdown table in this section is malformed (extra leading | on each row: || ...). This will render incorrectly in most Markdown viewers; use a standard 2-column table format with single leading/trailing pipes (e.g., | Task | Tool | / | --- | --- |).

Copilot uses AI. Check for mistakes.
Comment thread .claude/skills/submitting-jobs/SKILL.md Outdated

## Storage resolution order

The script must reference an S3 location. Resolve where it lives or will be uploaded in this order:
Copy link

Copilot AI Apr 6, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This section says the script "must reference an S3 location", but later the command reference explicitly allows a local .py file path and describes auto-upload. Consider rewording to clarify that runs execute from S3, and the CLI will upload local scripts automatically (unless --no-upload is used with an s3:// URI).

Suggested change
The script must reference an S3 location. Resolve where it lives or will be uploaded in this order:
Wherobots job runs execute from an S3 location, but the CLI accepts either a local `.py` file path or an `s3://` URI for `<script>`. If you pass a local file, the CLI resolves where to upload it in this order; if you pass an `s3://` URI, it uses that location directly (typically with `--no-upload`).

Copilot uses AI. Check for mistakes.
@kadolor kadolor marked this pull request as ready for review April 6, 2026 23:03
@kadolor kadolor requested a review from a team as a code owner April 6, 2026 23:03
@kadolor kadolor requested review from ClayMav and haizhou-zhao April 6, 2026 23:03
@sfishel18
Copy link
Copy Markdown

@kadolor i think we should try to be more minimalist with this skill content. a couple of reasons:

  1. primarily, the more tokens we add to the context window as a result of this skill the more we degrade the effectiveness of the llm and/or trigger compaction loops
  2. secondarily, due to the dynamic nature of the commands after wherobots api ..., we should avoid direct references to those because they could become wrong

imo the goal here should be more about "meta knowledge", focusing on:

  • teaching the agent how to use the cli in general (how to discover the individual commands, use the available global flags, etc)
  • common multi-step workflows like uploading a job script file and then running it.
  • guidance to the agent about when to use the cli, specifically as compared to using the mcp server instead

we shouldn't include any information that the agent/llm can easily find for itself through basic interactions with the cli

similar to what we discussed in the bug bash, it would be ideal to have a set of agentic tasks that we want the cli to enable. we could then run those tasks with and without the skill loaded. the skill should only contain content that makes the agent more effective in accomplishing the tasks

@ClayMav
Copy link
Copy Markdown
Member

ClayMav commented Apr 7, 2026

We need to place our skills in a different place so if you don't mind I am going to take another crack at this. I will send it to you @kadolor and @sfishel18 for review once I have something. It has to be this way since we will be distributing skills through multiple channels and a claude plugins and that is the common pattern

@kadolor
Copy link
Copy Markdown
Author

kadolor commented Apr 7, 2026

Closing this out, see conversation above

@kadolor kadolor closed this Apr 7, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Development

Successfully merging this pull request may close these issues.

4 participants