chore: add LLM behavior constraints to SKILL.md

Prevent the LLM from reasoning about internal query/polling logic,
attempting fallback strategies, or retrying on its own.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
ywkj 2026-03-20 07:40:51 +08:00
parent 81b932a97d
commit 5d42a64615
1 changed files with 7 additions and 0 deletions

View File

@ -42,6 +42,13 @@ Returns structured JSON with product data:
- Variants/SKUs
- Supplier info
## Rules — MUST follow
1. **Execute only, do not reason about internals.** Run the CLI command and return the output. Do NOT attempt to understand, debug, or bypass the scrape/query/polling logic inside the tool.
2. **No fallback strategies.** If the command fails, report the error as-is. Do NOT try alternative approaches such as browser scraping, direct HTTP calls, or different API endpoints.
3. **No retry loops.** If authentication or the scrape call fails, report the failure. Do NOT re-check configs, re-acquire tokens, or retry on your own.
4. **Trust the tool's output.** The CLI handles session management, retries, and error formatting internally. Your job is to invoke it and relay the result — nothing more.
## Reference
See [references/1688-product-master.md](references/1688-product-master.md).