
software-engineering
Evaluate software architecture and plan large changes.
- version
- 0.1.9
- GitHub stars
- 347
Install
/plugin install software-engineering@khanelinix-ai-toolsRun it inside Claude Code after adding the marketplace with /plugin marketplace add khaneliman/khanelinix.
Contents
Discovered from sourceSkills 1
Software Engineering
Use this skill for architecture decisions. Let repository canon define local policy, and let language, framework, security, Git, and delivery skills own specialist work.
Start from Evidence
- Read contributor canon and every instruction file governing target paths.
- State mode, desired outcome, constraints, stakeholders, and consequence of failure. Separate observed facts, inferences, assumptions, and unknowns.
- Inspect only enough repository breadth to trace relevant entry points, dependencies, state, side effects, tests, deployment, and prior decisions.
- Recover current behavioral contract before proposing structure. Treat code, tests, public interfaces, schemas, and operational evidence as potentially conflicting evidence; identify conflict instead of silently choosing.
- Load matching domain skills before judging version-sensitive APIs or implementation details. If none applies, inspect manifests and current code, then verify against primary official documentation through available live tools; state evidence gap when verification is unavailable.
Route One Primary Mode
- current-state repository or subsystem evaluation: audit existing architecture, maintainability, correctness, operability, and change readiness. Read repository-evaluation.md.
- large-change planning: convert a feature, refactor, migration, or technical-debt outcome into contracts, state changes, boundaries, increments, rollout, and verification. Read large-change-planning.md.
- proposed standalone software architecture review: challenge a design proposal or ADR before implementation and outside commit/PR/diff review. Read architecture-review.md.
Read foundations.md when choosing or explaining cross-cutting engineering principles.
For secondary Git-review support, do not select a primary mode or produce a
separate verdict. Apply root lenses to supplied architecture premises and return
them to $git-toolkit; the review contract in the premise-review method of
engineering-principles keeps the finding format and verdict. Use
$planning-with-files only when task state should persist across sessions or
compaction. Use $ai-tools-architect for AI configuration. Use security skills
for explicit security audits or threat models.
Apply Engineering Lenses Proportionally
- Behavior: inputs, outputs, errors, side effects, compatibility, and observable guarantees.
- State: abstract value, valid states, invariants, ownership, transitions, mutability, persistence, and recovery.
- Boundaries: cohesion, coupling, information hiding, dependency direction, naming, authority, and failure containment.
- Execution: ordering, concurrency, idempotency, partial failure, retries, timeouts, resource bounds, observability, and security where relevant.
- Evolution: expected change, migration, rollback, test seams, maintenance ownership, and removal path.
- Responsibility: affected stakeholders, harm, fairness, privacy, accessibility, accountability, and human control in proportion to impact.
Increase depth for public APIs or schemas, stored data, authentication or trust boundaries, concurrent or remote effects, or destructive operations. Increase depth for safety-, legal-, or livelihood-relevant workflows. Otherwise sample representative paths and state evidence limits.
Do not mechanically require ADTs, immutability, SOLID patterns, hierarchical names, adapters, distributed machinery, or approval gates. Introduce a boundary only when it enforces an invariant, contains failure, isolates likely change, clarifies ownership, or creates a needed test seam. Prefer smallest design that satisfies observed requirements and keeps an explicit path for credible change.
Produce Decision-Ready Output
- Lead with outcome and highest-risk evidence.
- Show current model before proposed delta when structure spans components.
- Tie each recommendation to requirement, evidence, invariant, failure mode, or measured change pressure.
- Include alternatives only where choice is material; explain rejected option and reversal cost.
- End with validation, rollout or follow-up, unresolved unknowns, and confidence.
Sponsors
Details
- Marketplace
- khaneliman/khanelinix
- Source
- modules/common/ai-tools/skills/software-engineering
- Author
- khaneliman
- Category
- Developer Tools
More from khaneliman/khanelinix
All 43- tddv0.1.4
- technical-writingv0.2.0
- to-specv0.1.1
- to-ticketsv0.1.1
- typescript-best-practicesv0.1.6
- unslopv0.1.7
- verification-harnessv0.1.5
- whyv0.1.6
- ai-tools-architectv0.1.21
- architectv0.1.7
- arenav0.1.5
- bevy-toolkitv0.1.6
- blast-radiusv0.1.6
- blender-toolkitv0.3.2
- develop-web-gamev0.1.7
- diagnosing-bugsv0.1.3
- dotnet-toolkitv0.1.2
- engineering-principlesv0.1.22
- engineering-workflowv0.1.27
- figure-it-outv0.1.10
- git-toolkitv0.1.15
- github-toolkitv0.1.33
- howv0.1.7
- interrogatev0.1.10