| name | evaluate-resource-auto |
| description | Approval-free variant of evaluate-resource for automated/batch use: reads a queued file from resources/tmp/, evaluates the URL, writes the review, and cleans up. |
| when_to_use | Use when invoked by the cron job to process a queued file from resources/tmp/. Not for interactive use — use evaluate-resource for manual evaluation instead. Follow all steps in order; do not shortcut based on this description. |
| argument-hint | [path to queued .md file in resources/tmp/] |
| model | opus |
| effort | high |
| allowed-tools | WebFetch Read Write Bash(gh *) Bash(rm *) Bash(mv *) Agent |
evaluate-resource-auto
Approval-free variant of evaluate-resource. Follow all steps from that skill with two differences:
-
Before step 1 — read the queued file at the path provided and extract the URL from the source or url frontmatter field. If neither is present or both are empty, append an ## Error section with the message "No URL found in frontmatter", move the file to resources/tmp/failed/, and stop.
-
Replace steps 5 and 6 — instead of presenting for approval, act on the agent's output directly:
DUPLICATE: insert a row at the top of the resources/log.md table (newest-first ordering, directly below the header), delete the source file, and stop.
EXTEND: append the ## Related entry to the existing review, insert a row at the top of resources/log.md, then delete the source file.
- Full draft: write to
resources/reviews/<slugified-title>.md, insert a row at the top of resources/log.md, then delete the source file.
If anything fails after the fetch, append an ## Error section describing what failed and move the source file to resources/tmp/failed/. Do not delete it.