engineering-fundamentals
Skill1 starsChanged 11 days agoMITView on GitHub
Define the universal engineering philosophy for all platform skills: discovery, contracts, anti-slop, quality gates. Never invoke directly. Do NOT use for platform-specific implementation.
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 8 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈8
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 22
No artifact. Consumed implicitly by all platform skills as a foundation reference.
- Distinctive: line 43
MANDATORY: Run before Phase 0 and before any file mutation.
- Very specific: line 45
Run bash scripts/pre-flight.sh and present state to user.
- Very specific: line 71
If no SPEC.md exists and this is new work (not a one-off tweak), invoke spec-driven-development.
- Very specific: line 89
Do NOT write code yet: Verify SPEC.md exists, invoke planning-and-task-breakdown.
- Very specific: line 93
Create design/DESIGN-LOCK.md (Direction, Palette, Typography, Decisions, References) + design/approved/.
- Very specific: line 109
Read guides/DESIGN-CORE.md for the universal Three Dials System (VARIANCE, MOTION, DENSITY), Brief Inference, color principles, and dark …
- Distinctive: line 113
Do not blend directions across platforms.
- Distinctive: line 123
Intentionality: NO defaults without justification.
- Distinctive: line 125
Tokens Over Hardcoding: NO hardcoded values. All from DESIGN.md/theme.
- Distinctive: line 127
Performance: NO convenience-over-performance tradeoffs.
- Distinctive: line 129
Accessibility: NOT optional. Always: 4.5:1 contrast, focus indicators, reduced motion.
- Very specific: line 177
Every platform skill follows guides/PLATFORM-SKILL-TEMPLATE.md — read it before creating.
- Distinctive: line 185
All platform skills include eval files (trigger, golden, adversarial) to verify correct activation and output quality.
≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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
Engineering Fundamentals
Foundation for all platform skills. Never invoke directly.
Platform skills add implementation specifics to this philosophy.
Output Contract
No artifact. Consumed implicitly by all platform skills as a foundation reference.
When to Use
Loaded by all platform skills. Never invoke directly.
When NOT to Use
- Platform-specific implementation (platform skills add details to this foundation)
Core Philosophy
- Contracts Before Code — No file until SPEC.md, DESIGN.md, DESIGN-LOCK.md, .gitignore exist.
- Discovery Before Design — No code until assumptions surfaced and confirmed.
- Anti-Slop — Default AI output is generic. Intentionality prevents that.
- Gates Before Progress — Every phase ends with user confirmation.
- Behavioral Discipline — See AGENTS.md Rule 0c. Think before coding. Simplicity first. Surgical changes. Goal-driven execution.
- Error Paths Are Main Paths — See below. Every tool call, gate, loop, and session needs a failure path designed at build time, not patched post-incident.
Pre-Flight: Repo State Check (Before ANY phase)
MANDATORY. Run before Phase 0 and before any file mutation.
Run bash scripts/pre-flight.sh and present state to user. Ask about branch intent before any action. See references/PRE-FLIGHT-DETAIL.md for full steps and state table.
Universal Process
Phase 0 — Language Detection
Spanish keywords → Spanish. English → English. Other → that language. Never mix.
Phase 1 — Discovery Gate (MANDATORY)
NO CODE UNTIL COMPLETE.
Surface: Audience, Core problem, Scale (MVP/Standard/Complex), Platform. Discover (min 5): Audience, Purpose, Scope, Context, Constraints. Extended: Data, Security, Integrations, Offline, Scalability.
Confirm: Summarize. Ask "Is this correct? Shall we proceed?" Only after explicit yes.
Phase 2 — Write Contracts
2A: SPEC.md
If no SPEC.md exists and this is new work (not a one-off tweak), invoke spec-driven-development. Must include: Objective, Scope, Tech stack, Project structure, Acceptance criteria, Boundaries.
2B: DESIGN.md — Visual Identity + Implementation Contract
DESIGN.md has two layers. See guides/DESIGN-MD-SCHEMA.md for the full 17-section template. See guides/DISCOVERY-GUIDE.md for the discovery interview flow that produces design/design-discovery.md.
Sections 1-12 (Visual Identity): Colors, Typography, Spacing, Border radius, Elevation, Motion, Component states. Filled by direction skills. Blocks if missing required fields.
Sections 13-17 (Implementation Contract): CSS Architecture, Theme Implementation, Component Framework, Platform Responsiveness, Asset Pipeline. Filled by platform DESIGN-GUIDE.md. Warns if missing (platform-specific).
What NEVER goes in (any section): Business logic, API routes, database schema, auth implementation, folder structure — those belong in SPEC.md or ADRs.
| State | Action |
| DESIGN.md exists | Read it, extract tokens. Validate against DESIGN-MD-SCHEMA.md. Build within. |
| No DESIGN.md, wants visual system | Run discovery, pick direction skill, generate 12-section visual contract. Platform fills sections 13-17. |
| No DESIGN.md, one-off task | Do task, mention DESIGN.md once. |
Do NOT write code yet. Verify SPEC.md exists, invoke planning-and-task-breakdown. Only after plan is confirmed, proceed to BUILD.
2C: Design Asset Lock (MANDATORY after visual approval)
Create design/DESIGN-LOCK.md (Direction, Palette, Typography, Decisions, References) + design/approved/. DESIGN-LOCK.md is a snapshot — never changes after approval. During BUILD, MUST re-read it before coding.
2D: Lifecycle Awareness
Respect AGENTS.md lifecycle:
- DEFINE →
spec-driven-development - PLAN →
planning-and-task-breakdown - BUILD → Platform-specific skill +
incremental-implementation - VERIFY →
debugging-and-error-recovery - REVIEW →
code-review-and-quality - SHIP →
shipping-and-launch
Phase 3 — Design Direction (Path B only)
Read guides/DESIGN-CORE.md for the universal Three Dials System (VARIANCE, MOTION, DENSITY), Brief Inference, color principles, and dark mode protocol.
Then apply the platform-specific skill's DESIGN-GUIDE.md for banned lists, font stacks, and tool-specific tokens.
Do not blend directions across platforms. Do not blend dial values.
Phase 4 — Anti-Slop Principles
Universal (read first): See guides/ANTI-SLOP-CORE.md for the universal AI tell catalogue, content density rules, copy protocol, and UI state requirements.
Platform-specific: Then apply the platform skill's ANTI-SLOP-GUIDE.md for banned fonts, platform-specific layout tells, and stack-specific rules.
Intentionality: NO defaults without justification. Every choice has a reason.
Tokens Over Hardcoding: NO hardcoded values. All from DESIGN.md/theme. Used twice = must be a token.
Performance: NO convenience-over-performance tradeoffs. Measure: 60fps, fast loads, minimal bundles.
Accessibility: NOT optional. Always: 4.5:1 contrast, focus indicators, reduced motion. Accessibility is a gate.
Phase 5 — Quality Gates
Universal (run first): See guides/PRE-FLIGHT-CORE.md for the universal pre-output checklist — color, content, motion, AI tells, accessibility, technical checks.
Platform-specific: Then run the platform skill's PRE-FLIGHT.md for build commands, layout verification, and native integration checks.
Before declaring complete:
- Contracts present — SPEC.md and DESIGN.md exist and were read.
- Design Lock present —
design/DESIGN-LOCK.mdexists and was read before coding. - No hardcoded tokens — All colors, fonts, spacing from token system.
- Accessibility — Contrast 4.5:1, focus indicators, reduced motion fallback.
- Build passes — Compilation succeeds without errors.
- Visual consistency — No design element deviates from DESIGN.md without explicit approval.
Phase 5b — Error Path Design
Inspired by Harness Books: "Error paths are main paths." Every tool call, gate, loop, and session needs a failure path. See references/ERROR-PATHS.md for the full failure path matrix and anti-patterns.
Phase 6 — Common Rationalizations
| Excuse | Reality | |---|---|---| | "Default for now." | #1 sign of AI slop. | | "Too small for DESIGN.md." | Even 3 tokens help one-offs. | | "Quality later." | Quality is a gate, not an afterthought. | | "User didn't ask." | Senior engineers apply best practices by default. | | "I remember design." | Context drifts. DESIGN-LOCK.md is ground truth. | | "I know what they want." | 1% confidence. Discovery forces 95%. |
Phase 7 — Red Flags
Code before requirements, assumptions unsurfaced, contracts skipped, hardcoded values, accessibility optional, DESIGN-LOCK.md stale, features not in spec.
Platform Skill Template
Every platform skill follows guides/PLATFORM-SKILL-TEMPLATE.md — read it before creating.
Phase 8 — Eval System Integration
All platform skills include eval files (trigger, golden, adversarial) to verify correct activation and output quality. The eval system follows Evaluation-Driven Development (EDD):
- Trigger tests — Does the skill fire when it should (and stay silent when it shouldn't)?
- Golden tests — Does the skill produce correct output against a rubric?
- Adversarial tests — Is the skill robust against rephrasing and edge cases?
→ See docs/EVAL-GUIDE.md for the complete eval system documentation.
→ Every generated skill must include eval cases (see skill-creator).
Verification
SPEC.md with Objective/Scope/Stack/Acceptance/Boundaries. DESIGN.md with visual tokens. DESIGN-LOCK.md read before coding. Discovery done before code. User confirmed each gate. No hardcoded tokens. Accessibility (4.5:1 contrast, focus, reduced motion). Build passes. Eval cases present and passing.
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- ku5ic/claude-kit★ 2same 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 repository59 other files it gives its agents.
- AGENTS.mdAGENTS.md
- templates/CLAUDE.mdCLAUDE.md
- adapt-skillSkill
- api-and-interface-designSkill
- architecture-analysisSkill
- audit-skillSkill
- backend-api-masterySkill
- browser-testing-with-devtoolsSkill
- ci-cd-and-automationSkill
- clarify-skillSkill
- cli-toolsSkill
- code-review-and-qualitySkill
- code-simplificationSkill
- context-engineeringSkill
- critique-skillSkill
- customize-opencodeSkill
- debugging-and-error-recoverySkill
- debugging-three-strikesSkill
- delight-skillSkill
- deprecation-and-migrationSkill
- dev-environment-auditSkill
- documentation-and-adrsSkill
- doubt-driven-developmentSkill
- frontend-desktopSkill
- frontend-mobileSkill
- frontend-pwaSkill
- frontend-ui-engineeringSkill
- frontend-webSkill
- fullstack-shippingSkill
- gateSkill
- git-init-and-versioningSkill
- git-workflow-and-versioningSkill
- hard-skillSkill
- idea-refineSkill
- incremental-implementationSkill
- industrial-brutalist-uiSkill
- interview-meSkill
- minimalist-uiSkill
- multi-agent-orchestrationSkill
- observability-and-instrumentationSkill
- optimize-skillSkill
- output-skillSkill
- performance-optimizationSkill
- planning-and-task-breakdownSkill
- polish-skillSkill
- project-health-checkSkill
- project-metricsSkill
- redesign-skillSkill
- security-and-hardeningSkill
- self-improvementSkill
- shipping-and-launchSkill
- skill-creatorSkill
- skill-improverSkill
- soft-premium-uiSkill
- source-driven-developmentSkill
- spec-driven-developmentSkill
- test-driven-developmentSkill
- typeset-skillSkill
- user-onboardingSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.

