[Hackathon] fix(test): flags documented "with --history / --run / --all" are silently ignored without that flag (their siblings exit 5) #376

Verified against (commit SHA or released version)

commit 988c0355bb05631e9c6c1785e443937ff2a88c0e (main, v0.14.0)

Discord Username / User ID

1149447580477042769

What does this improvement do?

Many options say in their own --help that they only work alongside another flag ("with --history: …", "with --run, …", "with --all: …"). The CLI already has a rule for these. When the enabling flag is missing it refuses with exit 5 and a nextAction naming the fix, instead of quietly doing something else. The test result action states the reason:

// `--env` narrows the history list; on the latest-result path it would be
// silently ignored (same rule as `test run --filter` without `--all`).

That rule is applied today to test result --env, test create/create-batch --env, test run --filter, all four of test rerun's --all flags (#160 → #163), testlist run --report/--gh-output/--summary-file, test run --gh-output/--summary-file and ci doctor --timeout. But 14 options documented the same way are still accepted without their enabling flag. They exit 0 and change nothing:

What a user or agent gets:

• test result test_x --since 24h prints the latest result, not the run history it asked for. Invalid values pass too: --page-size 0 and --since garbage exit 0.
• test create --plan-from plan.json --wait (forgot --run) creates the test, runs nothing, and exits 0, the same exit code as a passing --run --wait. A CI step written that way goes green without running a test. Same for create-batch … --wait.
• test run test_x --allow-empty exits 0 whatever the flag says, while test rerun test_x --allow-empty is refused with exit 5.

Reproduction (offline, --dry-run; plan.json is any valid FE plan, dir/ holds a copy):

# refused today: exit 5
testsprite test result test_x --env staging --dry-run --output json; echo "exit=$?"
testsprite test create --plan-from plan.json --env staging --dry-run --output json; echo "exit=$?"
testsprite test rerun test_x --allow-empty --dry-run --output json; echo "exit=$?"
# accepted and ignored: exit 0
testsprite test result test_x --since 24h --dry-run --output json; echo "exit=$?"
testsprite test result test_x --page-size 0 --dry-run --output json; echo "exit=$?"
testsprite test create --plan-from plan.json --wait --timeout 30 --dry-run --output json; echo "exit=$?"
testsprite test create-batch --plan-from-dir dir --max-concurrency 5 --dry-run --output json; echo "exit=$?"
testsprite test run test_x --allow-empty --dry-run --output json; echo "exit=$?"

Observed: 5, 5, 5, then 0, 0, 0, 0, 0. Each ignored invocation prints exactly what it prints without the flag. In the real (non-dry-run) path the checks sit in the command actions, before any request, so the result is the same: in runCreate, opts.wait is read only inside if (opts.run === true).

Expected. Each of the 14 is refused like its siblings: exit 5, VALIDATION_ERROR, details.field set to the flag, and a nextAction such as "--since only applies with --history (it filters the run list). Add --history, or remove --since."

Details / implementation notes

• Where. Every guard goes in front of existing ones of the same shape, using the existing localValidationError(field, message):
  • test result action (src/commands/test.ts, next to the --env check): the other with --history: flags. --rerun/--no-rerun and --no-header need command.getOptionValueSource(…) === 'cli', because Commander gives negatable flags a default.
  • runCreate / runCreateBatch: --wait/--timeout (and --max-concurrency in the batch) when run !== true, next to the existing --env requires --run.
  • test run action: --allow-empty without --all, next to --filter only applies with --all, with the wording test rerun already uses.
• Deliberately not included: test run --timeout and testlist run --timeout without --wait. Both commands can poll without --wait (an automatic environment tunnel), so --timeout is live there even though their help says "with --wait".
• Tests: one case per flag in the existing spec of each command (test.test.ts / the create and run specs), asserting exit 5 and details.field, plus a control showing that the same flag with its enabling flag is still accepted. Each fails on main today.
• Optional, to keep this from coming back: a table-driven test that walks the Commander tree, collects every option whose description starts with with --X, and checks it is refused without --X. The table would carry an explicit allow-list for the two live --timeout cases above.
• Compatibility: a script that passes one of these flags without its enabling flag gets exit 5 instead of 0. That is the same trade fix(rerun): reject explicit IDs with --all and --status/--skip-terminal without --all #163 made for test rerun, and the nextAction names the fix.

I already hold #372–#374, so this is free for anyone to pick up once it's triaged.

Confirmations

• I have searched existing issues and this is not a duplicate.
• I have provided my Discord identity above for reward coordination.