Fix monitor-ci.sh doc/script mismatch: no --commit flag exists

[jnasbyupgrade]

Summary

CLAUDE.md and the /ci skill's own SKILL.md described a --commit SHA
flag for monitor-ci.sh, but the script only ever took positional args
([repos] [branch] [sha_pgxntool_test] [sha_pgxntool]). That flag text
appears to have crept in from ../ai/CLAUDE.md's general gh CLI advice
(gh run list/gh pr checks do take --commit) getting conflated with
this script's own interface.

Concretely, an agent following the doc literally invoked
monitor-ci.sh --commit SHA, which silently misparsed into
REPOS="--commit" (falls through to the "both" default case) with the
SHA landing in BRANCH — the script then polled a nonexistent branch
for ~35 minutes before failing.

Changes

• CLAUDE.md: describe the actual positional interface instead of a
  --commit flag.
• .claude/skills/ci/SKILL.md: same fix, plus a separate pre-existing bug
  in the same file — the "most reliable" usage example was missing the
  leading repos arg and had the two SHA args in the wrong order
  relative to the script's real sha_pgxntool_test/sha_pgxntool order.
• .claude/skills/ci/scripts/monitor-ci.sh: added a guard that fails fast
  with a usage message on a --flag-style invocation, instead of silently
  misparsing and hanging until timeout — a defense against this recurring
  even if some other doc/agent assumes flag syntax again.

Why not add a --commit flag instead

Considered making the script match the (wrong) documented interface
instead of fixing the docs, per this project's usual preference for
fixing the more ergonomic direction. Decided against it:

• both mode — the default — needs two separate SHAs for two separate
  repos' pushes, so a single --commit SHA flag can't express the
  primary use case at all.
• The positional form is already load-bearing: the /release skill
  invokes monitor-ci.sh positionally (.claude/skills/release/SKILL.md),
  and SKILL.md's own documented 4-arg form already matches the script
  correctly (only its shorthand one-liner was wrong).

Test plan

• bash -n syntax check on the modified script
• Ran monitor-ci.sh --commit abc123def — confirmed it now fails
  fast with exit 64 and a clear usage message (the failure path this
  bug caused)
• Ran monitor-ci.sh pgxntool-test <nonexistent-branch> (normal
  positional form) — confirmed it proceeds past the new guard and
  enters the real monitor loop unchanged (the success path)
• Checked for existing test coverage of these files (CLAUDE.md,
  SKILL.md, monitor-ci.sh) — none exists; they're agent-facing
  docs/scripts outside the BATS test suite's scope

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Advanced

Run ID: 05cd5f73-4952-4842-9f70-5af2472f5331

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}