bulk

claude-routines:bulk

Fan out an operation across multiple files. Reports per-file status, stops on hard failures, continues on per-file 4xx.

Common bulk requests

Phrasing	Action
"deploy all" or "deploy everything"	Iterate routines/**/*.md (and any other configured location), run deploy on each
"deploy /" or "deploy everything in "	Iterate <dir>/**/*.md
"deploy all routines using "	Grep for the include directive across .md files; deploy each match
"set enabled:false on all in "	Edit each file's frontmatter (set enabled: false), then deploy each
"validate all"	Run validate on every .md file in known routine locations
"diff all"	Run diff on every file with a trigger_id
"dry-run all" or "preview deploy all"	Same fanout as deploy, but no API writes

Output format

One line per file with status. Examples:

OK   routines/morning-brain-digest.md      (updated; cron unchanged, prompt +18 chars)
OK   routines/norway-news-digest.md        (created; trigger_id trig_01ABC...)
SKIP routines/draft-experimental.md        (validate failed: missing env_id)
ERR  routines/oslo-apartment-hunter.md     (HTTP 400: cron interval too short)
OK   personal/finn-tesla-scan.md           (no change)

5 files: 3 OK, 1 skipped, 1 error.

Behavior on errors

• Stop the entire bulk on hard failures (HTTP 5xx, auth failure, file-system error). Report what's done, what failed, and abort.
• Continue on per-file 4xx (validation error, cron interval too short, etc.). Skip that file, log the reason, move on.
• Always validate before deploying. A single invalid file should NOT abort the whole bulk; just report and skip.

Bulk dry-run

For dry-run deploy all or preview deploy all: same fan-out, but no API write calls. Per-file output shows which operation each would be (CREATE vs UPDATE) and which fields would change.

See also

• deploy <file> for one routine at a time
• validate <file> to check one routine
• diff <file> to compare one routine