adversarial-code-review-pass-integrity
Skill11 starsChanged 49 days agoApache-2.0View on GitHub
Reviews a diff through the lens of data integrity — multi-item mutations, atomicity, ordering, idempotency. Use only when the adversarial-code-review coordinator spawns it, alongside the security and schema passes.
In this repository83 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- about-konductorSkill
- adr-generatorSkill
- adversarial-code-review-pass-schemaSkill
- adversarial-code-review-pass-securitySkill
- adversarial-code-reviewSkill
- adversarial-design-reviewSkill
- agents-md-authoringSkill
- app-discoverySkill
- architecture-advisorSkill
- architecture-diagram-generationSkill
- argumentation-referenceSkill
- asana-sprint-planningSkill
- asdlc-aspect-reviewSkill
- asdlc-code-simplifierSkill
- aws-service-validatorSkill
- backend-developmentSkill
- backend-reviewSkill
- claude-teams-behaviorSkill
- cmux-dispatchSkill
- code-reviewSkill
- constraintsSkill
- cost-estimationSkill
- cypress-test-implementationSkill
- decision-researchSkill
- decision-writingSkill
- delegation-protocolSkill
- deliberation-panelSkill
- design-doc-guidelinesSkill
- design-evaluationSkill
- design-impact-reviewSkill
- design-quality-checkSkill
- doc-accuracy-analyzerSkill
- document-formatsSkill
- dom-inspectionSkill
- dynamodb-designSkill
- dynamodb-validationSkill
- e2e-test-strategySkill
- external-researchSkill
- find-aws-skillsSkill
- frontend-developmentSkill
- frontend-reviewSkill
- git-mergeSkill
- git-workflowSkill
- historical-issues-registrySkill
- humanize-writingSkill
- iam-policy-designSkill
- infra-validationSkill
- kiro-design-generationSkill
- kiro-requirements-generationSkill
- kiro-task-generationSkill
- legacy-to-agentic-estimateSkill
- mux-dispatchSkill
- non-functional-requirementsSkill
- persistent-memorySkill
- plan-reviewSkill
- playwright-test-implementationSkill
- policy-validationSkill
- pre-planning-analysisSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
3 rules. About 0 are common, 2 are distinctive, and 1 is 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.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 29
Delete or overwrite state on an error path where rollback or compensation is required
- Very specific: line 35
Do not flag a single-item write as requiring a transaction — transactions apply to multi-item mutations that must succeed or fail together, …
- Distinctive: line 40
Before flagging a missing implementation, check whether the codebase already provides it.
≈ “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
Adversarial Code Review — Data Integrity Pass
Overview
Reviews a diff through the lens of data integrity and failure modes. This is one of three parallel review passes spawned by an adversarial-review coordinator SOP. The framing is neutral by design: review through the lens of integrity, do not assume any specific issue exists.
Usage
Spawned by an adversarial-review coordinator SOP (such as k-adversarial-pull-request-review) as one of three parallel subagent passes. May also be run directly against a diff.
What to look for
Review through the lens of data integrity. In this diff, does anything:
- Perform a multi-item mutation without a transaction — two or more writes that must succeed or fail together
- Issue an unconditional
putorupdateon a resource that can see concurrent writes - Miss a condition expression, version check, or optimistic-lock guard on a contested resource
- Introduce a write path that has no test exercising the mutation
- Retry a non-idempotent operation without an idempotency key
- Depend on ordering across independent async events without a sequencing mechanism
- Lose the return value of a write and then re-read the item redundantly
- Delete or overwrite state on an error path where rollback or compensation is required
For each concern, name the file:line, the concrete failure scenario (partial write, lost update, ordering race), and a specific fix.
Out of scope
- Do not flag a single-item write as requiring a transaction — transactions apply to multi-item mutations that must succeed or fail together, not to a lone
put/update. - Do not flag bulk or unconditional writes in test setup/teardown code (e.g. seeding fixture data before a test, tearing down state after) — this is expected test scaffolding, not a defect. This exempts setup/teardown specifically, not test files as a whole: a non-atomic write in the code path actually under test (the one the test's assertions target) is still in scope and must be flagged.
Codebase awareness
Before flagging a missing implementation, check whether the codebase already provides it. Flag reimplementation or bypass, not absence:
- Existing transaction or batch-write helpers → flag reimplementation
- Established idempotency-key middleware → flag routes that skip it
- Shared repository or DAO layer → flag direct client calls that bypass it
Checker handoff
Findings produced by this pass are candidates, not verdicts. The coordinator forwards them to a different-persona checker that applies adversarial-code-review in Validator Mode before anything reaches the CR. Produce your honest read.
Severity guidance
- CRITICAL — non-atomic multi-item mutation without a transaction; unconditional overwrite on a contested resource; delete/overwrite on an error path with no rollback
- IMPORTANT — write path without test coverage; retry of a non-idempotent operation without an idempotency key; missing condition expression that a shared pattern in this repo uses
- SUGGESTION — extra query where a write's return value could be used; condition expression present but weaker than warranted (e.g.
attribute_existswhere a version match would fit)
Output format
[SEVERITY] Data Integrity — file:line
Problem: {what is wrong and why it is a risk}
Fix: {concrete suggested change}Return the findings list. Do not deduplicate against other passes — the coordinator does that.
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.

