| name | loadstone-cli |
| description | Use the Loadstone CLI to search the RuneScape 3 Wiki for quests or item information and RuneMetrics for character information like levels, status, quests completion. Use when the user asks for RS3 wiki data, Ironman-focused answers, RuneMetrics stats, or when they want CLI command guidance for Loadstone. |
| metadata | {"author":"loadstone","version":"1.0"} |
Loadstone CLI Skill
Use this skill to operate the Loadstone CLI in this repository and return data from the RuneScape 3 Wiki or RuneMetrics in a user-friendly format.
When to use
- User asks for RS3 Wiki data (pages, categories, tables, training info)
- User asks for RuneMetrics profile or quest info
- User asks how to run or interpret Loadstone CLI commands
- User wants structured JSON output for LLM consumption
Required context
- Prefer
--json when the user wants structured output.
Command usage
loadstone <command> [args] [options]
Profile data workflow
For token-efficient profile queries, use the filtering options:
- Use
--skills-only for skill level queries (e.g., "what's my crafting level?") - reduces ~14k tokens to ~300
- Use
--completed-quests-only for completed quest counts
- Use
--started-quests-only for current quest progress
- Use
--not-started-quests-only for available quests to start
- Use
--quests-only for all quests with minimal data
- Add
--include-raw if you need the complete raw API response
Note: The raw field is excluded by default in JSON mode to save tokens. Use --include-raw only when necessary.
Examples:
loadstone profile "Some Player" --json --skills-only
loadstone profile "Some Player" --json --completed-quests-only
loadstone profile "Some Player" --json --started-quests-only
Page content workflow
- First request available headings with
--headings.
- Choose only the headings relevant to the user request.
- Request those sections with
--fields "Heading 1","Heading 2".
Examples:
loadstone page "Abyssal whip" --headings
loadstone page "Abyssal whip" --fields "Usage","Drop sources" --json
Output handling
- If the user asks for data, prefer
--json and return structured results.
- If the user asks for a summary, parse JSON and provide a concise summary.
- If a command fails, return the error and offer a fix (missing command, network, rate limit).
