standards-review
Skill1 starsChanged 24 days agoMITView on GitHub
Audit a diff against the standards the repo's own CLAUDE.md states - abstraction, comments, register, boundaries, naming. Use before a commit, when reviewing a branch or PR, or when the user asks whether a change follows the standards.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- Cosmian/kmstheir own version, first seen Oct 8, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- Cosmian/kms★ 372same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository6 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- bug-red-testSkill
- docs-syncSkill
- init-workflowSkill
- plan-startSkill
- task-runSkill
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 12
Read it first. A rule it does not state is not a finding here - do not import your own taste, and do not carry a rule over from another …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 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
standards-review
Correctness bugs are /code-review's job. This is the other axis: does the code
read like the rest of the repo.
The repo's CLAUDE.md owns the standards; this skill owns the procedure.
Read it first. A rule it does not state is not a finding here - do not import
your own taste, and do not carry a rule over from another repo.
Scope
git diff <base>...HEAD by default; the working tree when there is no base. Only
changed lines - a standard broken on an untouched line is not this review's find,
except a stale comment on a line you touched, which is always in scope to delete.
Checks worth running
Apply the ones the repo's standards actually state, in the form they state them.
| Check | What fails it |
| Escape hatches from the type system | A cast or suppression where an annotation, a guard or narrowing would do. A third-party unknown with no practical guard earns a typed wrapper, proposed - not a cast slipped in |
| Premature abstraction | A helper, component, constant or file with exactly one caller. Inline it. Being pinnable in isolation is not reuse - pin the branches through the caller that owns them. The wrapper a single caller does earn is a third-party primitive wrapped in the repo's own shape |
| Comment bar | The default is no comment. One survives only if it records a decision the code cannot express and a competent reader would otherwise get wrong - an ordering that looks arbitrary, a guard against a race, an undocumented external constraint. Never one per enum member, object key or function step. First reach is the rename, every time |
| Chat-narrative leak | "as requested", "as we discussed", "let's", pasted prompt text; "previously X now Y", "no longer used", "an earlier rule said". Also any reference to a gitignored working doc - "see TASKS.md", "TASK-3" - in code, comments, commit messages or test names |
| Dependency direction | An import that reaches back up the layering, or a write to state another layer owns. If lint does not already catch it, the finding names the lint rule that would |
| Immutable by default | A reassigned accumulator a map/filter/reduce expresses. Sequential async that must short-circuit uses recursion, not a mutable loop counter |
| Derive once | The same collection filtered twice, or a throwaway count recomputed from an array already in hand |
| Boundary guards | Data crossing a trust boundary - persisted state, a cross-context message, an API envelope - used without a runtime guard. The guard lives beside the type and runs at the boundary that widened the value, never on a literal the compiler already knows |
| User-facing copy | Punctuation or wording that breaks the repo's stated convention for strings a user reads |
| Test naming | The suite names the unit; the case states the behaviour in lowercase with no "should" and no implementation detail |
Test changes
This check runs whatever the repo's CLAUDE.md says, because a weakened test
hides a broken behaviour from every later reviewer and gate. It applies to every
changed test: a modified or removed assertion, a deleted or skipped case, an edited
fixture, a mock, a timeout, a snapshot.
For each one, decide which it is:
- Follows a behaviour change. The production side of the diff changes the behaviour this test pins. The new expectation states the newly intended behaviour, traceable to the task or the requirement, and pins it at least as tightly as before.
- Bent to pass. The expectation moved to match whatever the code now does, with no intended behaviour change behind it - or it pins less than it did.
Signs that a test was bent:
- An assertion loosened: an exact match turned partial, a value turned
toBeDefinedor truthy, a count turned "at least". - An expected value swapped for the observed one, with no production change that explains the new value.
- A case deleted, skipped,
.only'd or marked fixme. Fewer rows in a table-driven case. - A mock widened until it swallows the call that failed. An error-path assertion removed.
- A timeout raised or retries added to quiet a failure. A snapshot regenerated wholesale.
- A fixture edited so the failing input no longer occurs.
- A title renamed so it no longer states what is checked.
For every bent change, say two things:
- Coverage lost. The behaviour, branch or input that is no longer pinned.
- Behaviour risk. Whether the production change behind it can break existing behaviour for a caller or a user, as a concrete scenario: "an upload with a refund now totals 4290 cents too high". If no risk can be named, say so.
These findings are ⚠️ and lead the report, ahead of any standards finding. They are an alert for the owner to decide on, never fixed silently by rewriting the test again. A test change that follows a behaviour change gets one line saying which change justifies it.
Reporting
Findings ranked most-severe first, each with the file, the line, and the concrete consequence. Any test-change alert comes first. Say plainly when nothing failed - a clean diff is a result, not a reason to invent a nit.
For a broad diff, delegate to the standards-reviewer agent - spawned with
model set to models.deep from .claude/workflow.json - and relay what it
found.
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.