lintro-review
Skill1 starsChanged 12 days agoMITView on GitHub
AI branch review using lintro's own `lintro review` — a standard local pre-push pass alongside the coderabbit and greptile CLIs, and the review of record when those are rate-limited or unavailable. Use when asked for a lintro review, an interim/fallback AI review, or as part of the default pre-push review set.
- Reads credentials
In this repository41 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- analyze-codeSkill
- analyze-projectSkill
- analyze-testsSkill
- audit-mergesSkill
- babysit-prSkill
- backlogSkill
- branchSkill
- coderabbitSkill
- commitSkill
- greptileSkill
- implement-issuesSkill
- issueSkill
- jiraSkill
- lintro-addSkill
- lintro-verifySkill
- lintSkill
- properizeSkill
- pr-raycastSkill
- prSkill
- raycastSkill
- rebaseSkill
- reconcileSkill
- scorecardSkill
- stand-ciSkill
- stand-generalSkill
- stand-odinSkill
- stand-pySkill
- stand-rustSkill
- stand-tsSkill
- sweep-prsSkill
- test-apiSkill
- test-shellSkill
- testSkill
- test-ui-qsfSkill
- test-uiSkill
- turbo-addSkill
- turbo-testSkill
- turbo-verifySkill
- which-prSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Run lintro's own AI diff review (lintro review) on the branch.
- Distinctive: line 108
Default pre-push flow runs all three where available:
- Distinctive: line 114
When CR/Greptile are rate-limited, do not wait for their reset — the lintro pass (plus whichever of the other two ran) is the pre-push …
- Very specific: line 127
If --transport cli fails on a flag the installed claude doesn't accept, that's live signal for the AI CLI-transport contract tests …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Lintro AI Review (standard local pass)
Run lintro's own AI diff review (lintro review) on the branch. Since 2026-08-02
(owner instruction) this is a standard third local pre-push pass alongside
coderabbit and greptile, not just a fallback — and when those two are rate-limited
or down, the lintro pass is the AI review of record for the push.
When to use
- As part of the default pre-push review set (
coderabbit+greptile+ this). - CodeRabbit and/or Greptile are rate-limited, erroring, or otherwise unavailable.
- You want an AI pass on the branch diff (or working tree) right now.
At final merge, bot threads that did get posted on the PR must still be triaged; a local lintro pass doesn't resolve GitHub threads.
Two correctness rules (do not skip)
- Use released lintro via
uvx, neveruv run lintro. The reviewer engine must be stable and independent of the working tree — especially when the branch is itself modifying lintro's AI/provider layer (you must not review half-migrated code with itself). Floor is>=0.94.7: that release carries both the CLI auth-mode detection (#1859, subscription OAuth works) and the CLI stdout error surfacing (#1836). - Enable AI locally behind a
skip-worktreeguard. Many repos (incl. py-lintro) ship.lintro-config.yamlwith noai:block orai.enabled: false. Flip it on locally in a way that cannot be staged or committed.
Steps
Enable AI locally without commit risk — only if the project config has AI disabled (
grep -A2 '^ai:' .lintro-config.yaml; skip this step if alreadyenabled: true/review: true):git update-index --skip-worktree .lintro-config.yamlThen merge this
ai:mapping into.lintro-config.yaml(append/merge the block — the rest of the project config stays as-is, so--with-lintkeeps working):ai: enabled: true review: true provider: cursor model: cursor-grok-4.5-high- Preflight the engine — Cursor + Grok 4.5 is the standard reviewer (owner
instruction, 2026-08-11):
- Provider is
cursor, model iscursor-grok-4.5-high(Cursor's Grok 4.5 id — confirm withcursor-agent modelsif it errors), via--transport cli. Verify the CLI exists withcursor-agent --version. - Workspace trust is required once per directory, or lintro's invocation dies
with "Workspace Trust Required". Grant it non-interactively from the repo/worktree
root:
cursor-agent -p 'Reply with exactly: ok' --trust --model cursor-grok-4.5-high --output-format text— trust persists for subsequent runs. - Fallbacks, in order, only when
cursor-agentis missing or erroring (say so in the report): logged-inclaudeCLI with--transport cliand noANTHROPIC_API_KEYin the environment (OAuth session, subscription billing; lintro ≥0.94.7 handles--baredetection); thenANTHROPIC_API_KEYwith either transport; thenOPENAI_API_KEYwithai.provider: openaiand--transport api. Each fallback also updates the step-1 yaml'sai.providerto the matching engine (e.g.anthropicfor Claude,openaifor OpenAI) and fixes or removesai.modelso the provider never seescursor-grok-4.5-high(removing it uses the provider default) — then rerunsuvx; the review command reads the config, not the fallback list. - No engine available → run the step-5 teardown first (it is mandatory whenever step 1 ran), then STOP and report which credential/binary is missing.
- Provider is
Run the review against the fresh base (a stale local
mainreviews the wrong diff), including lint:base="$(git symbolic-ref --quiet --short refs/remotes/origin/HEAD 2>/dev/null | sed 's@^origin/@@' || echo main)" git fetch --quiet origin "$base" uvx --from 'lintro[ai]>=0.94.7' lintro review --base "origin/$base" --with-lint --transport cli --timeout 600- Swap
--transport cliforapiper the step-2 outcome. CLI-transport turns run the whole review in oneclaudeinvocation — pass--timeout 600(the 60s default is API-stream-sized and will kill a CLI turn; #1900). - Omitting
--depth/--strictnessuses the project's configured review defaults. For a heavier final pass, add--strictness thorough --depth 3(more spend).
- Swap
- Triage — fix real findings; note any you dismiss and why, as you would in a CR/Greptile thread.
Tear down the local toggle so it can never land (un-skip first, then restore):
git update-index --no-skip-worktree .lintro-config.yaml git checkout .lintro-config.yaml
Relationship to coderabbit / greptile
Default pre-push flow runs all three where available:
commit → [greptile ‖ coderabbit ‖ lintro-review] → prWhen CR/Greptile are rate-limited, do not wait for their reset — the lintro pass (plus whichever of the other two ran) is the pre-push review of record. Still triage any bot threads that get posted on the PR afterward.
Notes
- Do not run the project's test suite while the step-1
ai.enabled: truetoggle is active. It leaks into config/doctor tests and fails them (a false red that looks like a code defect). Tear the toggle down (step 5) before running any tests, or don't run tests during a review. - Review exit code 2 means "no review was produced" (missing credential, provider unreachable, quota); the JSON error envelope names the kind. Since 0.94.1 the CLI's stdout error detail is surfaced — read it before diagnosing.
- If
--transport clifails on a flag the installedclaudedoesn't accept, that's live signal for the AI CLI-transport contract tests (py-lintro #1614) — report it.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

