The Problem Skill | Agent Skills

Codex compatibility note:

Invoke repository skills with $skill-name in Codex; this mirrored copy rewrites legacy Claude /skill-name references.

Task tracker mandate: BEFORE executing any workflow or skill step, create/update task tracking for all steps and keep it synchronized as progress changes.

User-question prompts mean to ask the user directly in Codex.

Ignore Claude-specific mode-switch instructions when they appear.

Strict execution contract: when a user explicitly invokes a skill, execute that skill protocol as written.

Subagent authorization: when a skill is user-invoked or AI-detected and its protocol requires subagents, that skill activation authorizes use of the required spawn_agent subagent(s) for that task.

Do not skip, reorder, or merge protocol steps unless the user explicitly approves the deviation first.

For workflow skills, execute each listed child-skill step explicitly and report step-by-step evidence.

If a required step/tool cannot run in this environment, stop and ask the user before adapting.

Codex Project-Reference Loading (No Hooks)

Codex uses static project-reference loading instead of runtime-injected project docs. When coding, planning, debugging, testing, or reviewing, open project docs explicitly using this routing.

Always read:

docs/project-config.json (project-specific paths, commands, modules, and workflow/test settings)
docs/project-reference/docs-index-reference.md (routes to the full docs/project-reference/* catalog)
docs/project-reference/lessons.md (always-on guardrails and anti-patterns)

Missing/stale context route: If docs/project-config.json, the docs index, lessons.md, CLAUDE.md, AGENTS.md, or any task-required reference doc is missing or stale, auto-run $project-init or the narrow setup route ($project-config, $docs-init, $scan-all, $scan --target=<key>, $claude-md-init) before ordinary project-specific work. If Codex mirrors or AGENTS.md are missing/stale, ask the user to run $sync-codex; do not auto-run it.

Situation-based docs:

Backend/CQRS/API/domain/entity changes: backend-patterns-reference.md, domain-entities-reference.md, project-structure-reference.md
Frontend/UI/styling/design-system: frontend-patterns-reference.md, scss-styling-guide.md, design-system/README.md
Spec authoring, docs/specs/ pathing, or TC format: feature-spec-reference.md, spec-system-reference.md, spec-principles.md
Behavior/public-contract changes or spec-test-code sync: workflow-spec-test-code-cycle-reference.md plus the spec docs above
Derived spec indexes/ERDs/reimplementation guides: spec-system-reference.md and source Feature Specs under docs/specs/
Integration test implementation/review: integration-test-reference.md
E2E test implementation/review: e2e-test-reference.md
Code review/audit work: code-review-rules.md plus domain docs above based on changed files

Do not read all docs blindly. Start from docs-index-reference.md, then open only relevant files for the task.

[BLOCKING] Execute skill steps in declared order. NEVER skip, reorder, or merge steps without explicit user approval. [BLOCKING] Before each step or sub-skill call, update task tracking: set in_progress when step starts, set completed when step ends. [BLOCKING] Every completed/skipped step MUST include brief evidence or explicit skip reason. [BLOCKING] If Task tools are unavailable, create and maintain an equivalent step-by-step plan tracker with the same status transitions.

Quick Summary

Goal: Build distinctive, craft-driven product interfaces (dashboards, admin panels, SaaS apps) that emerge from domain exploration rather than AI defaults.

Workflow:

Intent — Identify who, what task, what feeling (ask user if unclear)
Domain Exploration — Produce 4 required outputs: domain concepts, color world, signature element, named defaults
Propose Direction — Present exploration + direction, get user buy-in
Build — Apply craft foundations (layering, tokens, typography, spacing)
Evaluate — Run mandate checks (swap, squint, signature, token tests) before showing

Key Rules:

Every design choice must have a WHY — "it's common" is not a reason
Produce 4 domain outputs before proposing any direction
Run 4 mandate checks before showing output to user
Save reusable patterns to docs/design-system/interface-system.md when 2+ uses
Component patterns: docs/project-reference/frontend-patterns-reference.md
Styling/BEM guide: docs/project-reference/scss-styling-guide.md

Scope:

Use for: Dashboards, admin panels, SaaS apps, tools, settings pages, data interfaces
Not for: Landing pages, marketing sites, campaigns — redirect to $frontend-design

The Problem

You will generate generic output. Your training has seen thousands of dashboards. The patterns are strong.

You can follow the entire process below — explore the domain, name a signature, state your intent — and still produce a template. Warm colors on cold structures. Friendly fonts on generic layouts. "Kitchen feel" that looks like every other app.

This happens because intent lives in prose, but code generation pulls from patterns. The gap between them is where defaults win.

The process below helps. But process alone doesn't guarantee craft. You have to catch yourself.

Where Defaults Hide

Defaults don't announce themselves. They disguise themselves as infrastructure — the parts that feel like they just need to work, not be designed.

Typography feels like a container. Pick something readable, move on. But typography isn't holding your design — it IS your design. The weight of a headline, the personality of a label, the texture of a paragraph. These shape how the product feels before anyone reads a word. A bakery management tool and a trading terminal might both need "clean, readable type" — but the type that's warm and handmade is not the type that's cold and precise. If you're reaching for your usual font, you're not designing.

Navigation feels like scaffolding. Build the sidebar, add the links, get to the real work. But navigation isn't around your product — it IS your product. Where you are, where you can go, what matters most. A page floating in space is a component demo, not software. The navigation teaches people how to think about the space they're in.

Data feels like presentation. You have numbers, show numbers. But a number on screen is not design. The question is: what does this number mean to the person looking at it? What will they do with it? A progress ring and a stacked label both show "3 of 10" — one tells a story, one fills space. If you're reaching for number-on-label, you're not designing.

Token names feel like implementation detail. But your CSS variables are design decisions. --ink and --parchment evoke a world. --gray-700 and --surface-2 evoke a template. Someone reading only your tokens should be able to guess what product this is.

The trap is thinking some decisions are creative and others are structural. no structural decisions. Everything is design. The moment you stop asking "why this?" is the moment defaults take over.

Intent First

Before touching code, answer these. Not in your head — out loud, to yourself or the user.

Who is this human? Not "users." The actual person. Where are they when they open this? What's on their mind? What did they do 5 minutes ago, what will they do 5 minutes after? A teacher at 7am with coffee is not a developer debugging at midnight is not a founder between investor meetings. Their world shapes the interface.

What must they accomplish? Not "use the dashboard." The verb. Grade these submissions. Find the broken deployment. Approve the payment. The answer determines what leads, what follows, what hides.

What should this feel like? Say it in words that mean something. "Clean and modern" means nothing — every AI says that. Warm like a notebook? Cold like a terminal? Dense like a trading floor? Calm like a reading app? The answer shapes color, type, spacing, density — everything.

If you cannot answer these with specifics, stop. Ask the user. Do not guess. Do not default.

Every Choice Must Be A Choice

For every decision, you must be able to explain WHY.

Why this layout and not another?
Why this color temperature?
Why this typeface?
Why this spacing scale?
Why this information hierarchy?

If your answer is "it's common" or "it's clean" or "it works" — you haven't chosen. You've defaulted. Defaults are invisible. Invisible choices compound into generic output.

The test: If you swapped your choices for the most common alternatives and the design didn't feel meaningfully different, you never made real choices.

Sameness Is Failure

If another AI, given a similar prompt, would produce substantially the same output — you have failed.

This is not about being different for its own sake. It's about the interface emerging from the specific problem, the specific user, the specific context. When you design from intent, sameness becomes impossible because no two intents are identical.

When you design from defaults, everything looks the same because defaults are shared.

Intent Must Be Systemic

Saying "warm" and using cold colors is not following through. Intent is not a label — it's a constraint that shapes every decision.

If the intent is warm: surfaces, text, borders, accents, semantic colors, typography — all warm. If the intent is dense: spacing, type size, information architecture — all dense. If the intent is calm: motion, contrast, color saturation — all calm.

Check your output against your stated intent. Does every token reinforce it? Or did you state an intent and then default anyway?

Product Domain Exploration

This is where defaults get caught — or don't.

Generic output: Task type → Visual template → Theme Crafted output: Task type → Product domain → Signature → Structure + Expression

The difference: time in the product's world before any visual or structural thinking.

Required Outputs

Produce all four outputs below before proposing any direction:

Domain: Concepts, metaphors, vocabulary from this product's world. Not features — territory. Minimum 5.

Color world: What colors exist naturally in this product's domain? Not "warm" or "cool" — go to the actual world. If this product were a physical space, what would you see? What colors belong there that don't belong elsewhere? List 5+.

Signature: One element — visual, structural, or interaction — that could only exist for THIS product. If you can't name one, keep exploring.

Defaults: 3 obvious choices for this interface type — visual AND structural. You can't avoid patterns you haven't named.

Proposal Requirements

Your direction must explicitly reference:

Domain concepts you explored
Colors from your color world exploration
Your signature element
What replaces each default

The test: Read your proposal. Remove the product name. Could someone identify what this is for? If not, it's generic. Explore deeper.

The Mandate

Before showing the user, look at what you made.

Ask yourself: "If they said this lacks craft, what would they mean?"

That thing you just thought of — fix it first.

Your first output is probably generic. That's normal. The work is catching it before the user has to.

The Checks

Run these against your output before presenting:

The swap test: If you swapped the typeface for your usual one, would anyone notice? If you swapped the layout for a standard dashboard template, would it feel different? The places where swapping wouldn't matter are the places you defaulted.
The squint test: Blur your eyes. Can you still perceive hierarchy? Is anything jumping out harshly? Craft whispers.
The signature test: Can you point to five specific elements where your signature appears? Not "the overall feel" — actual components. A signature you can't locate doesn't exist.
The token test: Read your CSS variables out loud. Do they sound like they belong to this product's world, or could they belong to any project?

If any check fails, iterate before showing.

Craft Foundations

Subtle Layering

This is the backbone of craft. Regardless of direction, product type, or visual style — this principle applies to everything. barely notice the system working. When you look at Vercel's dashboard, you don't think "nice borders." You just understand the structure. The craft is invisible — that's how you know it's working.

Surface Elevation

Surfaces stack. A dropdown sits above a card which sits above the page. Build a numbered system — base, then increasing elevation levels. In dark mode, higher elevation = slightly lighter. In light mode, higher elevation = slightly lighter or uses shadow.

Each jump should be only a few percentage points of lightness. You can barely see the difference in isolation. But when surfaces stack, the hierarchy emerges. Whisper-quiet shifts that you feel rather than see.

Key decisions:

Sidebars: Same background as canvas, not different. Different colors fragment the visual space into "sidebar world" and "content world." A subtle border is enough separation.
Dropdowns: One level above their parent surface. If both share the same level, the dropdown blends into the card and layering is lost.
Inputs: Slightly darker than their surroundings, not lighter. Inputs are "inset" — they receive content. A darker background signals "type here" without heavy borders.

Borders

Borders should disappear when you're not looking for them, but be findable when you need structure. Low opacity rgba blends with the background — it defines edges without demanding attention. Solid hex borders look harsh in comparison.

Build a progression — not all borders are equal. Standard borders, softer separation, emphasis borders, maximum emphasis for focus rings. Match intensity to the importance of the boundary.

The squint test: Blur your eyes at the interface. still perceive hierarchy — what's above what, where sections divide. But nothing should jump out. No harsh lines. No jarring color shifts. Just quiet structure.

This separates professional interfaces from amateur ones. Get this wrong and nothing else matters.

Infinite Expression

Every pattern has infinite expressions. No interface should look the same.

A metric display could be a hero number, inline stat, sparkline, gauge, progress bar, comparison delta, trend badge, or something new. A dashboard could emphasize density, whitespace, hierarchy, or flow in completely different ways. Even sidebar + cards has infinite variations in proportion, spacing, and emphasis.

Before building, ask:

What's the ONE thing users do most here?
What products solve similar problems brilliantly? Study them.
Why would this interface feel designed for its purpose, not templated?

NEVER produce identical output. Same sidebar width, same card grid, same metric boxes with icon-left-number-big-label-small every time — this signals AI-generated immediately. It's forgettable.

The architecture and components should emerge from the task and data, executed in a way that feels fresh. Linear's cards don't look like Notion's. Vercel's metrics don't look like Stripe's. Same concepts, infinite expressions.

Color Lives Somewhere

Every product exists in a world. That world has colors.

Before you reach for a palette, spend time in the product's world. What would you see if you walked into the physical version of this space? What materials? What light? What objects?

Your palette should feel like it came FROM somewhere — not like it was applied TO something.

Beyond Warm and Cold: Temperature is one axis. Is this quiet or loud? Dense or spacious? Serious or playful? Geometric or organic? A trading terminal and a meditation app are both "focused" — completely different kinds of focus. Find the specific quality, not the generic label.

Color Carries Meaning: Gray builds structure. Color communicates — status, action, emphasis, identity. Unmotivated color is noise. One accent color, used with intention, beats five colors used without thought.

Before Writing Each Component

Every time you write UI code — even small additions — state:

Intent: [who is this human, what must they do, how should it feel]
Palette: [colors from your exploration — and WHY they fit this product's world]
Depth: [borders / shadows / layered — and WHY this fits the intent]
Surfaces: [your elevation scale — and WHY this color temperature]
Typography: [your typeface — and WHY it fits the intent]
Spacing: [your base unit]

This checkpoint is mandatory. It forces you to connect every technical choice back to intent.

If you can't explain WHY for each choice, you're defaulting. Stop and think.

Design Principles

Token Architecture

Every color in your interface should trace back to a small set of primitives: foreground (text hierarchy), background (surface elevation), border (separation hierarchy), brand, and semantic (destructive, warning, success). No random hex values — everything maps to primitives.

Text Hierarchy

Don't just have "text" and "gray text." Build four levels — primary, secondary, tertiary, muted. Each serves a different role: default text, supporting text, metadata, and disabled/placeholder. Use all four consistently. If you're only using two, your hierarchy is too flat.

Border Progression

Borders aren't binary. Build a scale that matches intensity to importance — standard separation, softer separation, emphasis, maximum emphasis. Not every boundary deserves the same weight.

Control Tokens

Form controls have specific needs. Don't reuse surface tokens — create dedicated ones for control backgrounds, control borders, and focus states. This lets you tune interactive elements independently from layout surfaces.

Spacing

Pick a base unit and stick to multiples. Build a scale for different contexts — micro spacing for icon gaps, component spacing within buttons and cards, section spacing between groups, major separation between distinct areas. Random values signal no system.

Padding

Keep it symmetrical. If one side has a value, others should match unless content naturally requires asymmetry.

Depth

Choose ONE approach and commit:

Borders-only — Clean, technical. For dense tools.
Subtle shadows — Soft lift. For approachable products.
Layered shadows — Premium, dimensional. For cards that need presence.
Surface color shifts — Background tints establish hierarchy without shadows.

Don't mix approaches.

Border Radius

Sharper feels technical. Rounder feels friendly. Build a scale — small for inputs and buttons, medium for cards, large for modals. Don't mix sharp and soft randomly.

Typography

Build distinct levels distinguishable at a glance. Headlines need weight and tight tracking for presence. Body needs comfortable weight for readability. Labels need medium weight that works at smaller sizes. Data needs monospace with tabular number spacing for alignment. Don't rely on size alone — combine size, weight, and letter-spacing.

Card Layouts

A metric card doesn't have to look like a plan card doesn't have to look like a settings card. Design each card's internal structure for its specific content — but keep the surface treatment consistent: same border weight, shadow depth, corner radius, padding scale.

Controls

Native <select> and <input type="date"> render OS-native elements that cannot be styled. Build custom components — trigger buttons with positioned dropdowns, calendar popovers, styled state management.

Iconography

Icons clarify, not decorate — if removing an icon loses no meaning, remove it. Choose one icon set and stick with it. Give standalone icons presence with subtle background containers.

Animation

Fast micro-interactions, smooth easing. Larger transitions can be slightly longer. Use deceleration easing. Avoid spring/bounce in professional interfaces.

States

Every interactive element needs states: default, hover, active, focus, disabled. Data needs states too: loading, empty, error. Missing states feel broken.

Navigation Context

Screens need grounding. A data table floating in space feels like a component demo, not a product. Include navigation showing where you are in the app, location indicators, and user context. When building sidebars, consider same background as main content with border separation rather than different colors.

Dark Mode

Dark interfaces have different needs. Shadows are less visible on dark backgrounds — lean on borders for definition. Semantic colors (success, warning, error) often need slight desaturation. The hierarchy system still applies, just with inverted values.

Avoid

Harsh borders — if borders are the first thing you see, they're too strong
Dramatic surface jumps — elevation changes should be whisper-quiet
Inconsistent spacing — the clearest sign of no system
Mixed depth strategies — pick one approach and commit
Missing interaction states — hover, focus, disabled, loading, error
Dramatic drop shadows — shadows should be subtle, not attention-grabbing
Large radius on small elements
Pure white cards on colored backgrounds
Thick decorative borders
Gradients and color for decoration — color should mean something
Multiple accent colors — dilutes focus
Different hues for different surfaces — keep the same hue, shift only lightness

Workflow

Communication

Be invisible. Don't announce modes or narrate process.

Never say: "I'm in ESTABLISH MODE", "Let me check system.md..."

Instead: Jump into work. State suggestions with reasoning.

Suggest + Ask

Lead with your exploration and recommendation, then confirm:

"Domain: [5+ concepts from the product's world]
Color world: [5+ colors that exist in this domain]
Signature: [one element unique to this product]
Rejecting: [default 1] → [alternative], [default 2] → [alternative], [default 3] → [alternative]

Direction: [approach that connects to the above]"

[Ask: "Does that direction feel right?"]

If Project Has system.md

Read docs/design-system/interface-system.md and apply. Decisions are made.

If No system.md

Explore domain — Produce all four required outputs
Propose — Direction must reference all four
Confirm — Get user buy-in
Build — Apply principles
Evaluate — Run the mandate checks before showing
Offer to save

After Completing a Task

When you finish building something, always offer to save:

"Want me to save these patterns for future sessions?"

If yes, write to docs/design-system/interface-system.md:

Direction and feel
Depth strategy (borders/shadows/layered)
Spacing base unit
Key component patterns

What to Save

Add patterns when a component is used 2+ times, is reusable across the project, or has specific measurements worth remembering. Don't save one-off components, temporary experiments, or variations better handled with props.

Consistency Checks

If system.md defines values, check against them: spacing on the defined grid, depth using the declared strategy throughout, colors from the defined palette, documented patterns reused instead of reinvented.

This compounds — each save makes future work faster and more consistent.

Deep Dives

For more detail on specific topics:

references/principles.md — Code examples, specific values, dark mode
references/validation.md — Memory management, when to update system.md
references/critique.md — Post-build craft critique protocol

Commands

$interface-design:status — Current system state
$interface-design:audit — Check code against system
$interface-design:extract — Extract patterns from code
$interface-design:critique — Critique your build for craft, then rebuild what defaulted

frontend-design — Marketing pages, landing pages, screenshot replication
ui-ux-pro-max — Searchable design DB (50 styles, 21 palettes, 50 font pairings)
design-spec — Create design specifications from requirements (incl. UX research notes, accessibility audit)
web-design-guidelines — WCAG 2.2, accessibility review

Next Steps

MANDATORY IMPORTANT MUST ATTENTION — NO EXCEPTIONS after completing this skill, you MUST ATTENTION use a direct user question to present these options. Do NOT skip because the task seems "simple" or "obvious" — the user decides:

"$web-design-guidelines (Recommended)" — Review UI for accessibility and best practices
"$workflow-review-changes" — Review all changes before commit
"Skip, continue manually" — user decides

[IMPORTANT] Use task tracking to break ALL work into small tasks BEFORE starting — including tasks for each file read. This prevents context loss from long files. For simple tasks, AI MUST ATTENTION ask user whether to skip.

AI Mistake Prevention — Failure modes to avoid on every task:

Re-read files after context changes. Context compaction, resume, or long-running work can make memory stale; verify current files before acting. Verify generated content against source evidence. AI hallucinates APIs, names, claims, and document facts. Check the relevant source before documenting or referencing. Check downstream references before deleting or renaming. Removing an artifact can stale docs, generated mirrors, configs, and callers; map references first. Trace the full impact chain after edits. Changing a definition can miss derived outputs and consumers. Follow the affected chain before declaring done. Verify ALL affected outputs, not just the first. One green check is not all green checks; validate every output surface the change can affect. Assume existing values are intentional — ask WHY before changing. Before changing a constant, limit, flag, wording, or pattern, read nearby context and history. Surface ambiguity before acting — don't pick silently. Multiple valid interpretations require an explicit question or stated assumption with risk. Keep shared guidance role-relevant. Universal guidance must help every receiving skill or agent; code-specific obligations belong only in code-specific protocols.

Critical Thinking Mindset — Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act. Anti-hallucination: Never present guess as fact — cite sources for every claim, admit uncertainty freely, self-check output for errors, cross-reference independently, stay skeptical of own confidence — certainty without evidence root of all hallucination.

MUST ATTENTION apply critical + sequential thinking — every claim needs appropriate traced evidence (file:line for repo/code claims; source URL or artifact section for research, product, content, and docs claims); confidence >80% to act, <60% DO NOT recommend. Anti-hallucination: never present guess as fact, admit uncertainty freely, cross-reference independently, stay skeptical of own confidence.

MUST ATTENTION apply AI mistake prevention — verify generated content against evidence, trace downstream references before deleting or renaming, verify all affected outputs, re-read files after context loss, and surface ambiguity before acting.

Prompt-Enhance Closing Anchors

IMPORTANT MUST ATTENTION follow declared step order for this skill; NEVER skip, reorder, or merge steps without explicit user approval IMPORTANT MUST ATTENTION for every step/sub-skill call: set in_progress before execution, set completed after execution IMPORTANT MUST ATTENTION every skipped step MUST include explicit reason; every completed step MUST include concise evidence IMPORTANT MUST ATTENTION if Task tools unavailable, maintain an equivalent step-by-step plan tracker with synchronized statuses

Closing Reminders

IMPORTANT MUST ATTENTION — Protocols in force (concise digest of the SYNC/shared blocks this skill carries):

AI Mistake Prevention: verify generated content against evidence, trace downstream references, verify all affected outputs, re-read after context loss, surface ambiguity.
Critical Thinking: traced file:line proof per claim; confidence >80% to act; never guess.
MANDATORY IMPORTANT MUST ATTENTION break work into small todo tasks using task tracking BEFORE starting
MANDATORY IMPORTANT MUST ATTENTION search codebase for 3+ similar patterns before creating new code
MANDATORY IMPORTANT MUST ATTENTION cite file:line evidence for every claim (confidence >80% to act)
MANDATORY IMPORTANT MUST ATTENTION add a final review todo task to verify work quality

[TASK-PLANNING] Before acting, analyze task scope and systematically break it into small todo tasks and sub-tasks using task tracking.

[IMPORTANT] Analyze how big the task is and break it into many small todo tasks systematically before starting — this is very important.

Hookless Prompt Protocol Mirror (Auto-Synced)

Source: .claude/.ck.json + .claude/skills/shared/sync-inline-versions.md (:full blocks) + .claude/scripts/lib/hookless-prompt-protocol.cjs

[WORKFLOW-EXECUTION-PROTOCOL] [BLOCKING] Workflow Execution Protocol — MANDATORY IMPORTANT MUST CRITICAL. Do not skip for any reason.

Generic portability boundary: Reusable skills and protocol text stay project-neutral; project-specific conventions are discovered from docs/project-config.json and docs/project-reference/. Apply shared AI-SDD from shared/sdd-artifact-contract.md. Read docs/project-config.json and docs/project-reference/docs-index-reference.md, then open the project reference docs named there. For spec, test-case, behavior-change, public-contract, or docs/specs/ work, route through the local spec docs named by the docs index: feature-spec-reference.md, spec-system-reference.md, spec-principles.md, and workflow-spec-test-code-cycle-reference.md when specs/tests/code must stay synchronized. If either file or a required reference doc is missing or stale, auto-run $project-init (or the narrow lower-level route such as $project-config, $docs-init, $scan-all, or $scan --target=<key>) before ordinary project-specific work. Any supported AI tool may execute when this shared context and local docs are available.

DETECT: If the prompt starts with an explicit slash skill/workflow command, execute it directly. Otherwise match the prompt against the workflow catalog and skill list.
ANALYZE: Choose the best option: execute directly, invoke a skill, activate a standard workflow, or compose a custom step combination.
AUTO-SELECT: Pick the best option yourself. Do not ask the user to choose between direct execution, skill, standard workflow, or custom workflow.
ACTIVATE: For a selected workflow, call $start-workflow <workflowId>; for a selected skill, invoke that skill; for a custom workflow, sequence custom steps directly; for direct execution, proceed with the task.
CREATE TASKS: task tracking for ALL workflow/skill/custom steps before execution when the selected path has multiple steps.
EXECUTE: Advance per the Workflow Step Advancement & Parallel Phases rule in your context instructions — model-driven; a sub-agent completion advances a step identically to an inline call; a parallel-phase group is an all-return barrier (advance only after ALL members return, never serialize it)

Shared AI-SDD Protocol Markers

Source: .claude/skills/shared/sync-inline-versions.md

SYNC:ai-sdd-artifact-contract

AI-SDD Artifact Contract — Shared spec-driven development rules stay portable and source-owned.

Keep reusable AI-SDD principles in .claude; put repository-specific paths, commands, owners, products, and formats in project config/reference docs.

Preserve cycle: spec -> plan -> tasks -> implement -> verify -> update spec/docs.

Trace every requirement or invariant through decision, task, TC/test, source evidence, and docs/spec update.

Treat code-to-spec extraction as reference-only until accepted by the canonical spec owner.

Any supported AI tool may plan, implement, review, or verify with synced context; using multiple tools is optional.

Update .claude source first, then sync generated mirrors; do not manually edit .agents, .codex, or AGENTS.md. — why: mirrors are generated artifacts; hand-edits are overwritten on the next sync

If docs/project-config.json, root instruction files, or a required project-reference doc is missing or stale, auto-run $project-init or the narrow lower-level route before ordinary project-specific work.

Active reference: shared/sdd-artifact-contract.md in the active skills root.

SYNC:ai-sdd-artifact-contract:reminder

MANDATORY Apply shared/sdd-artifact-contract.md; keep reusable AI-SDD in .claude and local rules in project docs.
MANDATORY Code-to-spec extraction is reference-only until canonical acceptance; any supported AI tool may execute with synced context.
MANDATORY Update .claude source before syncing generated mirrors; do not manually edit .agents, .codex, or AGENTS.md.
MANDATORY Missing or stale project config, root instruction files, or required reference docs route project-specific work through $project-init or the narrow setup route automatically. [TASK-PLANNING] [MANDATORY] BEFORE executing any workflow or skill step, create/update task tracking for all planned steps, then keep it synchronized as each step starts/completes.

[LESSON-LEARNED-REMINDER] [BLOCKING] Task Planning & Continuous Improvement — MANDATORY. Do not skip.

Break work into small tasks (task tracking) before starting. Add final task: "Analyze AI mistakes & lessons learned".

Extract lessons — ROOT CAUSE ONLY, not symptom fixes:

Name the FAILURE MODE (reasoning/assumption failure), not symptom — "assumed API existed without reading source" not "used wrong enum value".
Generality test: does this failure mode apply to ≥3 contexts/codebases? If not, abstract one level up.
Write as a universal rule — strip project-specific names/paths/classes. Useful on any codebase.
Consolidate: multiple mistakes sharing one failure mode → ONE lesson.
Recurrence gate: "Would this recur in future session WITHOUT this reminder?" — No → skip $learn.
Auto-fix gate: "Could $code-review/$code-simplifier/$security-review/$lint catch this?" — Yes → improve review skill instead.
BOTH gates pass → ask user to run $learn. [CRITICAL-THINKING-MINDSET] Apply critical thinking, sequential thinking. Every claim needs traced proof, confidence >80% to act. Anti-hallucination principle: Never present guess as fact — cite sources for every claim, admit uncertainty freely, self-check output for errors, cross-reference independently, stay skeptical of own confidence — certainty without evidence root of all hallucination. AI Attention principle (Primacy-Recency): Put the 3 most critical rules at both top and bottom of long prompts/protocols so instruction adherence survives long context windows. Goal-driven execution: Define success criteria first, loop until verified, and stop only when observable checks pass. Tests verify intent: Tests must protect business rules/invariants and fail when the protected intent breaks, not only mirror current behavior.

Common AI Mistake Prevention (System Lessons)

Re-read files after context compaction. Edit requires prior Read in same context; compaction wipes read state. Re-read before editing.
Grep for old terms after bulk replacements. AI over-trusts find/replace completeness. Grep full repo after bulk edits for missed refs in docs/configs/catalogs.
Check downstream references before deleting. Deletions cascade doc/code staleness. Map referencing files before removal.
After memory loss, check existing state before creating new. Compaction wipes prior-work memory. Query current state to resume — never blindly duplicate.
Verify AI-generated content against actual code. AI hallucinates APIs, class names, method signatures. Grep to confirm existence before documenting/referencing.
Trace full dependency chain after edits. Changing a definition misses downstream consumers. Trace the full chain.
When renaming, grep ALL consumer file types. Some file types silently ignore missing refs (no compile error). Search code, templates, configs, generated files.
Trace ALL code paths when verifying correctness. Code existing ≠ code executing. Trace early exits, error branches, conditional skips — not just happy path.
Update docs that embed canonical data when source changes. Docs inlining derived data (workflows, schemas, configs) go stale silently. Update all embedding docs alongside source.
Verify sub-agent results after context recovery. Background agents may finish while parent compacted — grep-verify output, don't trust assumed completion.
Cross-check full target list against sub-agent assignments. Parallel sub-agents by category miss boundary items. Reconcile union of assignments against target list before proceeding.
Sub-agents inherit knowledge only from their agent .md definition — use custom agent types, not built-in Explore. Tool adoption = permission + knowledge + enforcement (numbered workflow step).
Persist sub-agent findings incrementally, not as a final batch. Long sub-agents hit cutoffs before final write — findings lost. Instruct append-per-section to report file.
When debugging, ask "whose responsibility?" before fixing. Trace caller (wrong data) vs callee (wrong handling). Fix at responsible layer — never patch symptom site.
Grep ALL removed names after extraction/refactoring. Primary file "done" ≠ secondary files clean. Grep entire scope for every removed symbol before declaring complete.
Assume existing values are intentional — ask WHY before changing. Pattern-matching as "wrong" skips context. Before changing any constant/limit/flag: read comments, git blame, surrounding code.
Verify ALL affected outputs, not just the first. One build green ≠ all green. Multi-stack changes (backend/frontend/tests/docs) require verifying EVERY output.
Evaluate fit before copying a nearby pattern. Closest example ≠ matching preconditions — verify the new context shares the same constraints, base classes, scope, lifetime.
Holistic-first debugging — resist nearest-attention trap. Don't dive into first plausible cause. List EVERY precondition (config, env vars, paths, DB, endpoints, creds, versions, DI, data). Verify each against evidence (grep/query — not reasoning). Ask "what would falsify this?" — if nothing, it's not a hypothesis. Most expensive failure: going deeper in "obvious" layer while bug sits in layer never questioned.
Surgical changes — apply the diff test (context-aware). Two modes: (1) Bug fix → every line traces to the bug; no restyling; orphan cleanup only for imports YOUR changes made unused. (2) Review/enhancement → implement improvements AND announce as "Enhancement beyond main request: [what]". Never silently scope-creep. Diff test: "Would this line exist if I wasn't asked to do X?" — if no, delete or announce.
Surface ambiguity before coding — don't pick silently. Multiple valid interpretations → present each with effort: "[Request] could mean (1) [N h], (2) [N h]. Which matters?" List scope/format/volume/constraints assumptions first. If simpler path exists, say so. Never silently pick.
[MANDATORY FIRST ACTION] ALWAYS activate a suitable skill or workflow BEFORE responding. Match task against workflow catalog + skill list; invoke via skill invocation or $start-workflow <workflowId>. NEVER answer or write code before checking. Skip = protocol violation.
Why-Review adversarial mindset — apply when reviewing any plan, decision, or design. Default SKEPTIC not VALIDATOR: steel-man a rejected alternative, invert each stated reason ("what does it sacrifice?"), stress-test top 2-3 assumptions, run pre-mortem ("ships, fails in 3 months — what breaks?"), surface 1-2 alternatives author missed. Section presence ≠ quality; quality = causal reasoning + concrete mitigations + evidence, not "it's better" or "monitor closely".
Front-load report-write in sub-agent prompts for large reviews. Many-file sub-agents hit budget before final write — findings lost. Design prompts so: (1) report-write is first explicit deliverable, (2) append per-file/section (not batched), (3) scope bounded so reads don't exhaust budget. Truncated mid-sentence with no report file → spawn narrower scope, don't retry same prompt.
After context compaction, re-verify all prior phase outcomes before continuing. Summaries describe intent, not environment state (git index, filesystem, processes). On resume, FIRST audit: git status, re-read modified files, verify filesystem. Every "completed" claim is an untested hypothesis until evidence confirms.
OOM/memory: check row count before row size. Triage: (1) Unbounded query — no DB filter for trigger? Push filter to DB; eliminates OOM. (2) Large rows? Projection reduces proportionally. Row reduction > projection in ROI.
Keep domain concepts out of generic/shared/infrastructure layers. Reusable layer (shared library, framework, infra module) must reference NO consumer-specific domain concept — tenant/customer/product IDs, business entities, feature rules. Leak compiles + runs → passes review silently while coupling the "reusable" layer to one consumer. Keep shared type domain-free; push domain fields/logic down into the consumer via subclass/composition. — why: a layer coupled to one consumer's domain is no longer reusable.

Agent Skills: The Problem

Install this agent skill to your local

Skill Files