[napetrov]

Summary

Item A3 of #94 §5 (hygiene), stage A. Refs #94.

hints printed to users
  personas discover   → "3. Approve: python cli.py personas approve …"   becomes  dokimi personas approve …
  personas approve    → "python cli.py questions generate …"            becomes  dokimi questions generate …
  baseline list       → "Run: python cli.py baseline save …"            becomes  dokimi baseline save …
  baseline compare    → "python cli.py baseline save …"                 becomes  dokimi baseline save …
report text (baseline-only limitation block)
  "python cli.py arms run --arms … [other parameters]"   becomes   dokimi arms run --product oneTBB --questions … --arms … --judge --judge-model …
commands/report.py
  cmd_report   removed (never registered with a parser; read args no parser defines); its import removed

The four hints already parsed with build_parser(); only the program name was stale, so flags are unchanged.
The arms run example did not parse (required --product/--questions missing, [other parameters] placeholder), so it gets concrete values.
The golden tests/fixtures/model_compare_sections.json is re-recorded; the diff is the four hint entries.

Numbers do not change.

Evidence

• Before: the new test fails on the parent commit, one failure per hint site:

$ pytest tests/test_printed_hints_parse.py
FAILED test_personas_discover_next_step_parses
FAILED test_personas_approve_next_step_parses
FAILED test_baseline_list_hint_parses
FAILED test_baseline_compare_hint_parses
FAILED test_baseline_only_report_hint_parses
E  AssertionError: hint names 'python', not the console script: python cli.py arms run --arms baseline,docs --judge --judge-model claude-sonnet-5-5 [other parameters]
E  assert 'python' == 'dokimi'
5 failed in 2.95s

• After: the same file passes, and the wider suite is green:

$ pytest tests/test_printed_hints_parse.py tests/test_runner_and_report.py tests/test_report_cli_inputs.py tests/test_baseline_compare.py tests/test_personas_approve.py tests/test_personas_discover_output.py tests/test_cli_help.py tests/test_cli_surface.py
240 passed in 14.02s

$ pytest            # full suite
9034 passed, 5 skipped in 333.57s

$ make ci-local CI_LOCAL_ARGS='--offline --jobs static,typecheck'
0 failed, 12 passed, 3 skipped   (ruff, schema, registry drift, growth, fragment, mypy, deptry, …)

Merge Danger

Door: two-way. Text-only change in printed hints and one unregistered function. Revert restores the old strings.

Blast radius: CLI output text (four hints, one report paragraph). No exit codes, flags, or artifact fields change. cmd_report had no caller, so removing it changes nothing reachable from the command line.

Anyone who copies an old hint from a log into a shell gets python cli.py …, which works only inside a source checkout. After this PR the hint prints the installed dokimi command.

Changelog fragment: changelog.d/20261026_130000_napetrov_cli-stale-hints.md (Fixed).

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.75%. Comparing base (b69b15f) to head (1984856).
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #252      +/-   ##
==========================================
+ Coverage   92.69%   92.75%   +0.05%     
==========================================
  Files         256      256              
  Lines       32421    32414       -7     
  Branches     5422     5421       -1     
==========================================
+ Hits        30054    30066      +12     
+ Misses       1579     1562      -17     
+ Partials      788      786       -2     

Files with missing lines	Coverage Δ
dokimi/dokimi/commands/baseline.py	93.05% <100.00%> (+3.70%)	⬆️
dokimi/dokimi/commands/personas.py	82.97% <100.00%> (+6.38%)	⬆️
dokimi/dokimi/commands/report.py	81.25% <ø> (+2.73%)	⬆️
dokimi/dokimi/report/model_compare.py	93.03% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[antonwolfy]

LGTM. The four hints and the baseline-only arms run example now name the dokimi console script, and the example parses. Removing cmd_report is safe.

What I ran, on Linux at 1984856:

• make ci-local: 26 passed, 2 failed, 7 skipped. The two failures are the gitleaks download (the proxy returned 504) and the commit scan that needs it, not this PR. Run by hand, gitleaks reports no leaks in the PR's commit.
• 9,036 tests passed, 3 skipped. Changed-line coverage is 100% (4 lines).
• Mutation run on the new test file, 16 mutants, 13 killed. Killed: each of the 5 sites reverted to python cli.py, a subcommand typo, --product dropped from the arms run example, [other parameters] restored, a wrong flag. The 3 survivors are abbreviated flags (--fil, --persona) and dropping the optional --product from baseline save; the hint still works when copied.
• With the base versions of baseline.py, personas.py and model_compare.py, all 5 new tests fail.

Checked, no issues

• cmd_report: no caller, registration, string dispatch or entry point refers to it. write_run_report and write_compare_report are imported from dokimi.report.markdown_report by their other users, not via commands/report.py. Ruff passes.
• arms run example: parsed with parse_args; it gives product=oneTBB, questions=questions/oneTBB.json, judge=True.
• Changelog fragment: present, one ### Fixed section, in the repo's format.