@felan-ai/design-system 0.1.2 → 0.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/README.md CHANGED
@@ -1,10 +1,10 @@
1
1
  # Felan AI — Design System
2
2
 
3
- > Copilots make engineers faster. Felan makes the engineering team faster.
3
+ > Delegate the work that shouldn't need anyone watching. Keep your attention for the hard problems.
4
4
 
5
- The design system for **Felan** — an agentic engineering platform. It captures the visual language, tone, and component vocabulary for Felan's marketing site (`felan.ai`) and product dashboard (`app.felan.ai`): canonical tokens and HTML/CSS component previews, plus a printable single-page catalog.
5
+ The design system for **Felan** — background AI agents for software engineering teams. It captures the visual language, tone, and component vocabulary for Felan's marketing site (`felan.ai`) and product dashboard (`app.felan.ai`): canonical tokens and HTML/CSS component previews, plus a printable single-page catalog.
6
6
 
7
- The full design spec — palette, typography, spacing, component anatomy, voice and copy rules — lives in [`project/README.md`](project/README.md). Agents start with the portable [`project/design.md`](project/design.md) guidance through the [`project/SKILL.md`](project/SKILL.md) skill.
7
+ The full design spec — palette, typography, spacing, component anatomy, voice and copy rules — lives in [`project/README.md`](project/README.md). Agents use the GitHub-installed [`felan-design`](skills/felan-design/SKILL.md) skill and its portable [`design.md`](skills/felan-design/design.md) guidance.
8
8
 
9
9
  ## Repository layout
10
10
 
@@ -12,17 +12,27 @@ The full design spec — palette, typography, spacing, component anatomy, voice
12
12
  ├── catalog.html ← single-page catalog: readme + every preview, print-ready
13
13
  ├── project/ ← the design system itself
14
14
  │ ├── README.md ← full design spec (start here)
15
- │ ├── design.md ← portable agent design guidance
16
- │ ├── SKILL.md ← skill contract for Claude Code / agents
17
15
  │ ├── tokens.css ← canonical tokens + semantic type classes
18
16
  │ ├── colors_and_type.css ← preview entry with hosted Geist fonts
19
17
  │ ├── styles.css ← font-free production entry
20
18
  │ ├── assets/ ← logos, icons, customer & integration marks
21
19
  │ ├── templates/ ← reusable customer-facing artifacts, including commercial quotes
22
20
  │ └── preview/ ← design-system cards (brand, colors, type, components…)
21
+ ├── skills/felan-design/ ← GitHub-installed agent skill + design.md
23
22
  └── scripts/repo-browser/ ← local HTTP server for the catalog
24
23
  ```
25
24
 
25
+ ## Agent skill
26
+
27
+ Install the two-file `felan-design` skill directly from GitHub:
28
+
29
+ ```sh
30
+ npx skills add felan-ai/design-system --skill felan-design
31
+ ```
32
+
33
+ The skill bundles only `SKILL.md` and `design.md`. Tokens, assets, templates,
34
+ and lint rules remain in the npm package.
35
+
26
36
  ## Browse the catalog
27
37
 
28
38
  The catalog fetches the readme, manifest, and previews at load time, so it must be served over HTTP (opening `catalog.html` from `file://` won't work):
@@ -35,14 +45,14 @@ scripts/repo-browser/stop.sh
35
45
 
36
46
  Or run it in the foreground with `scripts/repo-browser/serve.sh`. By default the OS assigns a free local port, so parallel repo browsers do not overlap. Pass a preferred starting port if needed: `scripts/repo-browser/start.sh <port>`. Requires Node ≥ 18, no dependencies.
37
47
 
38
- The catalog mirrors the Claude Design view of this system: the project readme first, then every design-system card grouped by section (Brand, Colors, Components, Spacing, Type).
48
+ The catalog renders the project readme first, then every preview grouped by section (Brand, Colors, Components, Spacing, Type).
39
49
 
40
50
  ## npm package
41
51
 
42
52
  This repository publishes `@felan-ai/design-system`, which contains the
43
53
  canonical CSS, tokens, Felan-owned assets, reusable quote template, design
44
- specification, portable agent guidance, agent skill, and shared `@shadcn/lint`
45
- policy.
54
+ specification, and shared `@shadcn/lint` policy. The agent skill is installed
55
+ separately from GitHub and is not part of the npm package.
46
56
 
47
57
  ```sh
48
58
  pnpm add @felan-ai/design-system
@@ -57,8 +67,7 @@ The package's HSL-triplet tokens are consumed with `hsl(var(--token))`.
57
67
  self-host Geist. Static previews use `colors_and_type.css`, which also loads
58
68
  Geist from Google Fonts. Public exports are documented in the package README
59
69
  and include `./styles.css`, `./tokens.css`, `./colors_and_type.css`, `./spec`,
60
- `./design`, `./skill`, `./assets/*`, `./templates/quote.html`, and
61
- `./eslint/shadcn`.
70
+ `./assets/*`, `./templates/quote.html`, and `./eslint/shadcn`.
62
71
 
63
72
  See [`docs/consuming.md`](docs/consuming.md) for the complete export table,
64
73
  package exclusions, lint limitations, asset policy, and consumer rollout.
@@ -80,6 +89,6 @@ Tailwind 4 migration. `go-to-market` consumes the assets and templates only.
80
89
  - The design medium is HTML/CSS/JS **prototypes** — reference implementations to match visually, not production code to copy structurally.
81
90
  - The reusable commercial quote template lives in `project/templates/quote/`; see its README for
82
91
  placeholders and print instructions.
83
- - `project/_ds_manifest.json` and `project/_ds_bundle.js` are generated by the design tooling; the catalog renders from the manifest, so regenerating it keeps the catalog in sync.
92
+ - The repo browser derives catalog metadata from each preview's `@dsCard` comment, so previews stay in sync without generated files.
84
93
  - Static previews load Geist Sans/Mono from the Google Fonts CDN; production consumers provide their own font files.
85
- - `npm pack --dry-run` and the package verifier are the source of truth for the published file surface; catalog previews and generated metadata are repository-only.
94
+ - `npm pack --dry-run` and the package verifier are the source of truth for the published file surface; catalog previews are repository-only.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@felan-ai/design-system",
3
- "version": "0.1.2",
3
+ "version": "0.2.0",
4
4
  "description": "Felan design tokens, brand guidance, reusable assets, templates, and lint policy",
5
5
  "license": "MIT",
6
6
  "type": "module",
@@ -24,8 +24,6 @@
24
24
  "project/tokens.css",
25
25
  "project/colors_and_type.css",
26
26
  "project/README.md",
27
- "project/design.md",
28
- "project/SKILL.md",
29
27
  "lint",
30
28
  "project/assets/felan-logo.svg",
31
29
  "project/assets/felan-logo.png",
@@ -45,8 +43,6 @@
45
43
  "./tokens.css": "./project/tokens.css",
46
44
  "./colors_and_type.css": "./project/colors_and_type.css",
47
45
  "./spec": "./project/README.md",
48
- "./design": "./project/design.md",
49
- "./skill": "./project/SKILL.md",
50
46
  "./lint-policy": "./lint/shadcn-policy.json",
51
47
  "./eslint/shadcn": "./lint/eslint-shadcn.mjs",
52
48
  "./assets/felan-logo.svg": "./project/assets/felan-logo.svg",
package/project/README.md CHANGED
@@ -1,7 +1,7 @@
1
1
  # Felan AI — Design System
2
2
 
3
- > **Copilots make engineers faster. Felan makes the engineering team faster.**
4
- > Felan is an agentic engineering platform — the team layer for AI-native software development: cloud background agents that work inside the tools your team already uses, share team knowledge, follow team guardrails, and improve from feedback.
3
+ > **Delegate the work that shouldn't need anyone watching. Keep your attention for the hard problems.**
4
+ > Background agents for engineering teams — your models, your infrastructure, no markup on tokens.
5
5
 
6
6
  This design system captures the visual language, tone, and component vocabulary for Felan's marketing site and product dashboard. Use it to design new surfaces, slides, mocks, and interfaces that feel unmistakably *Felan*.
7
7
 
@@ -9,11 +9,14 @@ This design system captures the visual language, tone, and component vocabulary
9
9
 
10
10
  ## Sources
11
11
 
12
- This system began as a reverse-engineering of the **Bugzy** (QA agent) codebase and brand, then had its palette migrated to Felan's pine-green and its copy/IA rewritten for the platform positioning. It is now the canonical **Felan platform** design system.
12
+ The visual system originated in the Bugzy web app (Next.js + Tailwind + shadcn/ui); treat that as provenance only. Canonical tokens live in `tokens.css`; the mark is `assets/felan-logo.svg` / `.png`.
13
13
 
14
- - **Visual origin:** the Bugzy web app (Next.js + Tailwind + shadcn/ui) — token structure, warm-neutral ramp, Geist type, orb/glow treatment, component anatomy all carried over and are approved as-is. Those legacy `web/` code paths are historical provenance only; do not treat them as Felan product spec.
15
- - **Brand:** `assets/felan-logo.svg` / `.png` — the forest-green geometric aperture mark.
16
- - **Canonical tokens:** `tokens.css`. All extracted assets live in `assets/`.
14
+ ## Name and brand world
15
+
16
+ - **Felan** is Irish for "wolf" (from *Faolán*). The grey wolf is the brand's creature.
17
+ - The brand world is the **taiga**: pine green on warm neutrals, northern, calm, deep.
18
+ - The wolf appears in illustration and character art outside product UI — never in the logo, which is the geometric aperture mark.
19
+ - The legal entity (Cloud Automation Solutions) appears only in legal documents; everywhere else the company is **Felan**.
17
20
 
18
21
  ---
19
22
 
@@ -21,54 +24,65 @@ This system began as a reverse-engineering of the **Bugzy** (QA agent) codebase
21
24
 
22
25
  Felan ships two primary surfaces, both represented in this system:
23
26
 
24
- 1. **Marketing site** (`felan.ai`) — dark, dense, gradient-glow hero, green brand pop, Geist typography. Section order follows the messaging hierarchy: **Hero** (individual AI gains → team-level delivery) → **Problem** (context, standards, follow-through, guardrails didn't scale with AI) → **Product** (the four differentiators) → **Integrations** → **Proof** (start with a repo; customer logos live here) → **Pricing** → **CTA / Footer**.
27
+ 1. **Marketing site** (`felan.ai`) — dark, dense, gradient-glow hero, green brand pop, Geist typography. Section order follows the messaging hierarchy: **Hero** (background agents for engineering teams; delegate the work) → **Problem** (developer attention is the bottleneck; agents that need someone watching don't scale) → **Use cases** (triggered by events / recurring work / started by a person) → **Why a managed platform** (always on, safe to leave alone, many at once, owned by the team) → **Cost and control** (the differentiators) → **Build vs. buy / Pricing** → **CTA / Footer**.
25
28
  2. **Product dashboard** (`app.felan.ai`) — light/dark responsive, sidebar shell grouped **Work** (Sessions, Activity, Approvals) / **Automate** (Automations, Skills) / **Team** (Agents, Memory) / **Configuration** (Integrations, Models & keys, Settings). Views: sessions, activity, approvals, automations, agents, integrations.
26
29
 
27
30
  ---
28
31
 
29
32
  ## Content fundamentals
30
33
 
31
- **What Felan is:** an agentic engineering platform — the team layer for AI-native software development. Engineers got faster with AI (Claude Code, Cursor, Copilot, Codex, Devin); the team layer around that work — shared context, standards, coordination, autonomy boundaries — did not keep up. Felan gives engineering orgs cloud background agents that work in existing tools, share team knowledge, follow guardrails, and improve from feedback.
34
+ **What Felan is:** background AI agents for software engineering teams. Agents start on their own — from an event, a schedule, or a person — work without anyone watching, and run on shared infrastructure (our cloud or the customer's) instead of developers' laptops. A designated person reviews the result at a set human-in-the-loop point.
35
+
36
+ **The anchor line:** *Delegate the work that shouldn't need anyone watching. Keep your attention for the hard problems.*
37
+ **The product line:** *Background agents for engineering teams — your models, your infrastructure, no markup on tokens.*
32
38
 
33
- **The anchor line:** *Copilots make engineers faster. Felan makes the engineering team faster.*
39
+ **Local agents and copilots stay.** Developers keep using them for the hard problems. Never position Felan against them; Felan takes the work that shouldn't need anyone watching.
34
40
 
35
- **The core distinction to teach on every surface** — copilot vs. background agent: one engineer → the team · a person prompts it → events/schedules/delegation trigger it · local chat context → shared team memory · personal autonomy → scoped permissions, approvals, audit logs · advice for a person → work advanced in Slack, tickets, Git, CI, docs.
41
+ **The hero is developer attention.** Cost-efficient background agents are the differentiator: no markup on tokens, any major model provider, full cost visibility, managed cloud or self-hosted.
36
42
 
37
- **Voice:** confident, technical, pragmatic. Calm authority, zero hype. Felan speaks like a senior engineer — it doesn't sell "AI," it says what changes for the team. No exclamation marks.
43
+ **Voice:** confident, technical, pragmatic. Calm authority, zero hype. Felan speaks like a senior engineer — it doesn't sell "AI," it says what work gets delegated and what it costs. No exclamation marks.
38
44
 
39
- **Audience:** engineering leaders (CTO / VP Eng / Head of Engineering / Head of Platform) as buyers, AI-forward senior/staff/platform engineers as champions. **Qualify by AI adoption, not headcount** — teams whose engineers adopted AI faster than the team's SDLC adapted. There is no company-size threshold and no QA framing.
45
+ **Audience:** software engineering teams already using coding agents. Buyers are engineering leaders (CTO / VP Eng / Head of Engineering / Head of Platform); champions are senior/staff/platform engineers. There is no company-size threshold.
40
46
 
41
- **POV:** second-person *you* for benefits; third-person *Felan* as a named teammate (Felan is both the platform and the lead agent). Avoids first-person "we" except in legal/about copy.
47
+ **POV:** second-person *you* for benefits; third-person *Felan* for the product. Avoid first-person "we" except in legal/about copy and founder content.
42
48
 
43
49
  **Casing:** Title Case for headlines, sentence case for subheads and UI labels. Short declarative section headers.
44
50
 
45
51
  **Sentence rhythm:** short, then explanatory.
46
- - Headline: `Copilots make engineers faster. Felan makes the team faster.`
47
- - Subhead: `Felan gives your engineering team cloud background agents that carry delegated work forward — with shared team knowledge and guardrails, inside the tools you already use.`
48
- - Feature title: `Runs in the background`
49
- - Feature subtitle: `Agents keep delegated work moving when nobody is prompting them.`
52
+ - Headline: `Background Agents for Engineering Teams`
53
+ - Subhead: `Delegate work to agents that start on their own. Your models, your infrastructure, no markup on tokens.`
54
+ - Feature title: `Runs while you're away`
55
+ - Feature subtitle: `Agents pick up incidents, failed builds, and dependency updates without anyone watching.`
50
56
 
51
57
  **Em-dashes** are a signature punctuation mark — they break up explanatory clauses naturally. Expect 1–2 per paragraph of body copy.
52
58
 
53
59
  **Emoji:** none in UI chrome, ever. No emoji in headlines, buttons, or empty states.
54
60
 
55
- **Numbers are concrete — but only claims that are structurally true from the brief.** Use: `Start with one repo`, `Works in Slack, GitHub, Linear, Jira`, `Every agent action logged`. Do **not** carry over Bugzy-era proof claims (`5-minute setup`, `No config required`, `SOC2 ready`, `EUR 250/month`, `dev/QA ratio >8:1`, `auto-healed`). Where an illustrative number is needed, mark it as sample copy (see the marketing kit's pricing note). Never invent customer quotes, metrics, or case-study outcomes.
61
+ **Claims:** only what the positioning supports:
62
+ - `Start with one repo`
63
+ - `No markup on tokens`
64
+ - `Bring your own keys or subscription`
65
+ - `All major model providers, including local models`
66
+ - `See what every agent and task costs`
67
+ - `Managed cloud or your own infrastructure`
68
+ - `Open-source agent` (Felan Code)
69
+ - Benchmark savings only with their test scope stated.
56
70
 
57
- **Categories / eyebrows:** `Agentic engineering platform`, `The gap`, `The product`, `Proof`, `Integrations`, `Pricing`. Short, sentence/title-cased pill above section H2s.
71
+ Never invent customer quotes, metrics, outcomes, prices, compliance claims, or setup times. Where an illustrative number is needed, mark it as sample copy.
58
72
 
59
- **Naming — hard rules:**
60
- - **Felan** is the platform *and* the lead agent — one name for both. Wordmark lockup stays **"Felan AI"** (from the domain `felan.ai`); in body copy the name is **Felan** alone. No new wordmark variants.
61
- - **Bugzy** is the QA agent — the first agent shipped, on its own site `bugzy.ai`. It is *one agent under the Felan canopy*, never the platform. QA may appear as **one example of delegated work among several** (a single row in a mock, an example low on the page) — never as the category, never in a headline.
62
- - Name agents by **SDLC role**: Developer agent, QA agent, SRE agent, General Purpose agent. No character names or mascots.
63
- - **Legal footer:** `© 2026 Felan`. No "GmbH," no Berlin/Munich location line — that was fabricated during extraction.
73
+ **Categories / eyebrows:** `Background agents`, `The problem`, `Use cases`, `Cost`, `Pricing`. Short, sentence/title-cased pill above section H2s.
64
74
 
65
- **FAQ answer example:** *"Does Felan replace engineers? No — humans own direction, judgment, and approvals. Felan carries the delegated work between those decisions."*
75
+ **Naming — hard rules:**
76
+ - **Felan** is the product and the company. Wordmark lockup stays **"Felan AI"** (from the domain `felan.ai`); in body copy the name is **Felan** alone. No new wordmark variants.
77
+ - **Felan Code** is the open-source coding agent at the core of the platform — part of Felan, not a separate product.
78
+ - Describe agents by the work they do ("an agent investigates the incident"). No named personas, characters, or mascots.
79
+ - **Legal footer:** `© 2026 Felan`.
66
80
 
67
- **Words Felan uses** (and you should too): *background agent, session, delegate / delegation, automation, trigger, guardrails, permissions, approvals, audit log, shared team knowledge, team memory, team layer, SDLC, repo, PR, rollout, context, follow-through.*
81
+ **FAQ answer example:** *"Does Felan replace my coding agent? No — keep using it for the hard problems. Felan takes the work that shouldn't need anyone watching."*
68
82
 
69
- **Retire the QA-era vocabulary** (only allowed inside an explicitly-QA example): *coverage gap, regression, triage, auto-heal, flaky, test maintenance, parallel runs, pass rate, "QA that keeps up".*
83
+ **Words Felan uses:** *background agent, delegate, unattended, human-in-the-loop, review, trigger, schedule, session, model choice, bring your own keys, cost per task, self-hosted, repo, PR.*
70
84
 
71
- **Keep avoiding, and now also avoid:** *revolutionary, seamless, cutting-edge, supercharge, unlock, game-changer, magical, solution, ecosystem, synergy, empowers, transforms, "10x"*, and any *"AI-powered / AI-driven / AI-first"* framing — assume the audience knows it's AI; say what changes for the team. Don't position Felan as an orchestrator of other vendors' agents (every agent is company-built), don't imply it replaces engineers (humans own direction, judgment, approvals), and don't lead with security/governance as the hero (guardrails prove adoptability; the hero is team-level productivity).
85
+ **Avoid:** *revolutionary, seamless, cutting-edge, supercharge, unlock, game-changer, magical, solution, ecosystem, synergy, empowers, transforms, "10x"*, and any *"AI-powered / AI-driven / AI-first"* framing. Say *delegate*, not *hand off*, *offload*, or *outsource* — developers use *offload* and *outsource* for giving up understanding. Don't position Felan as an orchestrator of other vendors' agents (Felan runs its own agent), don't imply it replaces engineers (humans own direction, judgment, and review), and don't compete with local agents or copilots.
72
86
 
73
87
  ---
74
88
 
@@ -79,8 +93,46 @@ Felan ships two primary surfaces, both represented in this system:
79
93
  - **Brand green — `hsl(152 44% 26%)`** is the single accent color. Used for: primary CTAs (`variant="brand"`), hero glow orbs, focus rings, chart-1, active states, sidebar accent. On dark backgrounds it brightens to `hsl(152 46% 34%)`.
80
94
  - **Warm neutrals.** The entire neutral ramp is tinted with hue 20–30 (warm) — backgrounds are `hsl(30 10% 95%)` (light) / `hsl(20 10% 7%)` (dark), never pure gray. This warmth is non-negotiable; it's what makes the site feel human rather than enterprise-sterile.
81
95
  - **Surface elevation:** three tinted layers (`--surface-1` → `--surface-3`) step up in luminance on light, down on dark. Cards sit on `surface-1`; featured cards sit on `surface-2`.
82
- - **Status colors:** green (`success`, done/approved), red (`destructive`, reject/high-risk), yellow (`warning`, awaiting approval), blue (`info`, running). In mockups: `/10` background and `/20` border for pills.
83
- - **Categorical chart palette:** green → teal → blue → amber → violet (5 colors, rotates on integration category cards and pricing tier top-bars).
96
+ - **Status colors:** `success` for done/approved, `destructive` for reject/high-risk, `warning` for awaiting approval, and `info` for running. Use labels or icons as well as color.
97
+ - **Categorical chart palette:** green, teal, blue, amber, violet. Use `chart-1` through `chart-5` for data visualization only, not product UI categories or text. Category identity in product UI uses neutral chips such as `bg-muted border text-foreground`; the icon and label identify the category.
98
+
99
+ ### Status colors
100
+
101
+ Canonical values are HSL triplets. Each status has four roles, because one color cannot serve as readable text and as a fill: yellow text that passes contrast on white is brown, and red text that passes on near-black is pink.
102
+
103
+ | Role | Token | Use |
104
+ | --- | --- | --- |
105
+ | Text | `--success` | Text, icons, status dots, borders at reduced opacity |
106
+ | Subtle | `--success-subtle` | Tinted background of pills, alerts and inline banners |
107
+ | Solid | `--success-solid` | Fill of buttons, solid badges and prominent banners |
108
+ | On solid | `--success-foreground` | Text and icons on the solid fill |
109
+
110
+ | Token | Light | Dark |
111
+ | --- | --- | --- |
112
+ | `success` | `142 72% 26%` | `142 69% 58%` |
113
+ | `success-subtle` | `140 60% 93%` | `142 45% 12%` |
114
+ | `success-solid` | `142 72% 29%` | `142 64% 30%` |
115
+ | `success-foreground` | `0 0% 100%` | `0 0% 98%` |
116
+ | `warning` | `42 100% 28%` | `45 93% 58%` |
117
+ | `warning-subtle` | `50 100% 90%` | `42 60% 12%` |
118
+ | `warning-solid` | `45 96% 53%` | `45 96% 53%` |
119
+ | `warning-foreground` | `20 14% 11%` | `20 10% 7%` |
120
+ | `info` | `214 84% 42%` | `213 94% 68%` |
121
+ | `info-subtle` | `214 100% 95%` | `214 55% 15%` |
122
+ | `info-solid` | `214 84% 45%` | `214 80% 47%` |
123
+ | `info-foreground` | `0 0% 100%` | `0 0% 98%` |
124
+ | `destructive` | `0 72% 44%` | `2 88% 64%` |
125
+ | `destructive-subtle` | `0 90% 96%` | `2 55% 14%` |
126
+ | `destructive-solid` | `0 72% 47%` | `1 72% 47%` |
127
+ | `destructive-foreground` | `0 0% 98%` | `0 0% 98%` |
128
+
129
+ Each status text color has at least 4.5:1 WCAG contrast on `background`, `card`, `muted`, and its own `-subtle` fill, in both modes. Each `-foreground` has at least 4.5:1 contrast on its `-solid` fill. The light warning solid is bright yellow at 1.6:1 against the page, so it always carries text; status dots use the text color.
130
+
131
+ - Text or icons use `text-success`.
132
+ - Pills and alerts use `bg-success-subtle border border-success/30 text-success`; keep body text neutral.
133
+ - Buttons, solid badges and prominent banners use `bg-success-solid text-success-foreground`.
134
+ - Status dots use `bg-success` with a visible label beside the dot.
135
+ - Substitute `warning`, `info`, or `destructive` for the other statuses. Let the tokens switch themes; do not add palette-based `dark:` overrides.
84
136
 
85
137
  ### Backgrounds — the signature glow
86
138
 
@@ -184,7 +236,6 @@ The marketing homepage is dark (`20 10% 7%`) with **two radial gradient orbs**:
184
236
 
185
237
  - Customer logos: SVG, displayed monochrome-to-color on hover.
186
238
  - Screenshots: minimal, mostly product UI — sessions, approvals, delegated work. No stock photography. No illustrated people. No 3D renders.
187
- - **Asset flags — replace before launch:** `flaky-hero.png`, `og-image.png`, and `github-app-feature-card.png` are **Bugzy-era QA product shots**. They must be replaced with Felan sessions/delegation product shots (a sessions view, an approval queue, agents working in Slack). Don't fake a screenshot — flagged here as a real replacement need.
188
239
 
189
240
  ### Layout rules
190
241
 
@@ -222,7 +273,7 @@ Recurring icons and their meanings:
222
273
  2. `svgl.app` for Linear / Framer-era logos
223
274
  3. Self-hosted SVG in `/public/integrations/*.svg` or `/public/customers/*.svg` for the rest
224
275
 
225
- These are all copied into `assets/integrations/` and `assets/customers/` here.
276
+ Integration marks are copied into `assets/integrations/` here.
226
277
 
227
278
  **Emoji:** not part of the system. The only emoji in the codebase (`👀 🎯 📁 📄`) appear inside mocked-up Slack messages or file-tree UI, never in real product chrome.
228
279
 
@@ -236,24 +287,20 @@ These are all copied into `assets/integrations/` and `assets/customers/` here.
236
287
 
237
288
  ```
238
289
  ├── README.md ← you are here
239
- ├── design.md ← portable guidance for agents building Felan interfaces
240
- ├── SKILL.md ← skill contract for Claude Code / agents
241
290
  ├── tokens.css ← canonical tokens + semantic type classes
242
291
  ├── colors_and_type.css ← preview entry with hosted Geist fonts
243
292
  ├── styles.css ← font-free production entry
244
- ├── assets/ ← logos, icons, customer & integration marks, avatars
293
+ ├── assets/ ← logos, icons, integration marks, avatars
245
294
  │ ├── felan-logo.png ← primary mark (forest-green “io” aperture on cream)
246
295
  │ ├── felan-logo.svg ← primary mark, vector (forest green, transparent)
247
296
  │ ├── felan-icon-192.png ← app icon
248
297
  │ ├── felan-icon-512.png ← app icon (hi-res)
249
298
  │ ├── felan-icon.svg ← vector mark
250
- │ ├── customers/ ← Invenda, NEG, Toki, GoodShape, SportAitchQ (logos only)
251
299
  │ ├── integrations/ ← Asana, ClickUp, Discord, Linear, etc.
252
300
  │ └── avatars/ ← headshots (legacy)
253
301
  ├── preview/ ← design-system cards (rendered in Design System tab)
254
- ├── templates/ ← reusable customer-facing artifacts
255
- │ └── quote/ ← A4 commercial quote HTML template + placeholder contract
256
- └── fonts/ ← (empty) — production consumers self-host Geist
302
+ └── templates/ ← reusable customer-facing artifacts
303
+ └── quote/ ← A4 commercial quote HTML template + placeholder contract
257
304
  ```
258
305
 
259
306
  ## Font substitution — flagged
@@ -77,7 +77,7 @@
77
77
  <div class="lockup"><img src="{{LOGO_SRC}}" alt="Felan AI"><span class="wordmark">Felan AI</span></div>
78
78
  <div class="meta">{{QUOTE_LABEL}}<br>{{REFERENCE}}<br>{{ISSUE_DATE}}</div>
79
79
  </header>
80
- <div class="eyebrow" style="margin-top: 22mm">Agentic engineering platform</div>
80
+ <div class="eyebrow" style="margin-top: 22mm">Background AI agents</div>
81
81
  <h1>{{HEADLINE}}</h1>
82
82
  <p class="lede">{{INTRO}}</p>
83
83
  <div class="recipient"><strong>Prepared for {{CUSTOMER}}</strong>{{CONTACT}}<br><span class="small">Prepared by {{SELLER}} · {{LEGAL_SELLER}}</span></div>
@@ -50,8 +50,10 @@
50
50
  --brand: 152 44% 26%; /* @kind color */
51
51
  --brand-foreground: 0 0% 100%; /* @kind color */
52
52
 
53
- /* Destructive */
54
- --destructive: 0 84% 60%; /* @kind color */
53
+ /* Destructive: text, subtle tint, solid fill, text on the solid fill */
54
+ --destructive: 0 72% 44%; /* @kind color */
55
+ --destructive-subtle: 0 90% 96%; /* @kind color */
56
+ --destructive-solid: 0 72% 47%; /* @kind color */
55
57
  --destructive-foreground: 0 0% 98%; /* @kind color */
56
58
 
57
59
  /* Structural */
@@ -72,10 +74,19 @@
72
74
  --chart-4: 45 85% 50%; /* @kind color */ /* amber */
73
75
  --chart-5: 280 60% 55%; /* @kind color */ /* violet */
74
76
 
75
- /* Status (marketing + product) */
76
- --success: 142 71% 45%; /* @kind color */
77
- --warning: 38 92% 50%; /* @kind color */
78
- --info: 210 90% 55%; /* @kind color */
77
+ /* Status (marketing + product): text, subtle tint, solid fill, text on the solid fill */
78
+ --success: 142 72% 26%; /* @kind color */
79
+ --success-subtle: 140 60% 93%; /* @kind color */
80
+ --success-solid: 142 72% 29%; /* @kind color */
81
+ --success-foreground: 0 0% 100%; /* @kind color */
82
+ --warning: 42 100% 28%; /* @kind color */
83
+ --warning-subtle: 50 100% 90%; /* @kind color */
84
+ --warning-solid: 45 96% 53%; /* @kind color */
85
+ --warning-foreground: 20 14% 11%; /* @kind color */
86
+ --info: 214 84% 42%; /* @kind color */
87
+ --info-subtle: 214 100% 95%; /* @kind color */
88
+ --info-solid: 214 84% 45%; /* @kind color */
89
+ --info-foreground: 0 0% 100%; /* @kind color */
79
90
 
80
91
  /* Sidebar (product app) */
81
92
  --sidebar-background: 30 8% 97%; /* @kind color */
@@ -166,9 +177,24 @@
166
177
  --brand: 152 46% 34%; /* @kind color */ /* brighter on dark */
167
178
  --brand-foreground: 0 0% 100%; /* @kind color */
168
179
 
169
- --destructive: 0 63% 31%; /* @kind color */
180
+ --destructive: 2 88% 64%; /* @kind color */
181
+ --destructive-subtle: 2 55% 14%; /* @kind color */
182
+ --destructive-solid: 1 72% 47%; /* @kind color */
170
183
  --destructive-foreground: 0 0% 98%; /* @kind color */
171
184
 
185
+ --success: 142 69% 58%; /* @kind color */
186
+ --success-subtle: 142 45% 12%; /* @kind color */
187
+ --success-solid: 142 64% 30%; /* @kind color */
188
+ --success-foreground: 0 0% 98%; /* @kind color */
189
+ --warning: 45 93% 58%; /* @kind color */
190
+ --warning-subtle: 42 60% 12%; /* @kind color */
191
+ --warning-solid: 45 96% 53%; /* @kind color */
192
+ --warning-foreground: 20 10% 7%; /* @kind color */
193
+ --info: 213 94% 68%; /* @kind color */
194
+ --info-subtle: 214 55% 15%; /* @kind color */
195
+ --info-solid: 214 80% 47%; /* @kind color */
196
+ --info-foreground: 0 0% 98%; /* @kind color */
197
+
172
198
  --border: 20 6% 18%; /* @kind color */
173
199
  --input: 20 6% 18%; /* @kind color */
174
200
  --ring: 152 46% 34%; /* @kind color */
package/project/SKILL.md DELETED
@@ -1,33 +0,0 @@
1
- ---
2
- name: felan-design
3
- description: Use whenever creating, changing, or reviewing a Felan interface, page, visual artifact, or customer-facing brand copy. Covers production UI, marketing pages, prototypes, documents, and presentations.
4
- user-invocable: true
5
- ---
6
-
7
- # Felan Design
8
-
9
- Before doing visual or brand-copy work, read `design.md` next to this file. It is the portable agent guidance bundled with this skill. Packaged consumers expose the same file as `@felan-ai/design-system/design`.
10
-
11
- Read `README.md` when the task needs the full component, asset, iconography, or content reference. Read `tokens.css` for exact token values. Inspect the consuming application's existing components before proposing new ones.
12
-
13
- ## Workflow
14
-
15
- 1. Identify the reader, task, surface, primary action, evidence, and required states.
16
- 2. Inspect the repository's framework, routes, components, tokens, and conventions.
17
- 3. Choose the marketing, product, or artifact guidance in `design.md`.
18
- 4. Reuse existing components and assets; add only what the task cannot express already.
19
- 5. Implement responsive and accessible behavior without introducing a parallel design system.
20
- 6. Run the repository's required checks and review the result against the pre-ship checklist in `design.md`.
21
-
22
- If essential product, audience, content, or delivery details are missing, ask focused questions before designing. If the user invokes this skill without a specific deliverable, ask what they want to build and whether it is production UI or a visual artifact.
23
-
24
- For static slides, mocks, or prototypes, copy the required packaged assets and produce a viewable artifact. For production work, preserve the application's architecture and use its installed design-system package and components.
25
-
26
- ## Included References
27
-
28
- - `design.md` — required portable guidance for agent decisions and verification
29
- - `README.md` — voice, tone, visual foundations, iconography, content fundamentals
30
- - `tokens.css` — canonical HSL tokens (light + dark) and semantic typography classes
31
- - `colors_and_type.css` — preview entry that adds hosted Geist fonts
32
- - `assets/` — packaged Felan marks, app icons, and integration SVGs
33
- - `preview/` — repository-only design cards for colors, type, spacing, components, and brand
package/project/design.md DELETED
@@ -1,118 +0,0 @@
1
- # Felan Design Guidance
2
-
3
- Use this guide whenever designing or implementing a Felan interface, page, prototype, document, or presentation.
4
-
5
- Felan is an agentic engineering platform — the team layer for AI-native software development. Felan is also the lead agent. Bugzy is one QA agent under Felan, never the platform.
6
-
7
- Anchor line: **Copilots make engineers faster. Felan makes the engineering team faster.** Qualify the audience by AI adoption, not company size.
8
-
9
- This is the portable agent guide. `README.md` contains the full reference, while `tokens.css` is authoritative for token values.
10
-
11
- ## Start With the Reader and Task
12
-
13
- Before choosing a layout, identify:
14
-
15
- 1. Who is using or reading this?
16
- 2. What must they understand, decide, approve, configure, or complete?
17
- 3. What is the single most important next action?
18
- 4. What evidence and states are needed to support that task?
19
-
20
- Shape the hierarchy around those answers. Do not start from a generic dashboard, landing-page template, or card grid. Do not invent features, proof, data, or infrastructure to make a design look complete.
21
-
22
- ## Choose the Correct Surface
23
-
24
- ### Marketing
25
-
26
- - Force dark mode.
27
- - Lead with team-level delivery, not AI novelty or governance.
28
- - Use soft green and blue radial orbs in the hero and final CTA; keep interior sections restrained.
29
- - Use one green CTA per screen. Render other actions as neutral, outline, ghost, or text links.
30
- - Show sessions, approvals, delegation, and work inside existing tools. Do not use stock photography, illustrated people, 3D renders, or fake product screenshots.
31
- - Add customer proof, prices, metrics, and logos only when approved content and assets exist.
32
-
33
- ### Product
34
-
35
- - Default to light mode and support dark mode.
36
- - Favor clear hierarchy, readable information density, and predictable interactions.
37
- - Reuse the established application shell, navigation, components, and state patterns.
38
- - Make status, ownership, next actions, approvals, and audit information easy to scan.
39
- - Reserve green for the primary action, active or focus state, and positive semantic state. Do not make every control green.
40
-
41
- ### Prototypes, Documents, and Presentations
42
-
43
- - Use the same tokens, typography, assets, voice, and claim rules.
44
- - Structure the artifact around its reader rather than imitating the marketing site or product shell.
45
- - Use an existing template when one fits. Do not create a parallel artifact framework for a one-off deliverable.
46
-
47
- ## Use the Existing System
48
-
49
- Inspect the consuming codebase before creating anything.
50
-
51
- - Reuse existing components and variants before adding new ones.
52
- - Preserve the host framework, routes, layout conventions, and interaction patterns.
53
- - Import `styles.css` for the font-free production foundation and use `hsl(var(--token))` values from `tokens.css`.
54
- - Use the consuming application's Geist installation. Hosted fonts in `colors_and_type.css` are for static previews.
55
- - Treat `preview/` as visual reference, not production structure to copy.
56
- - Add a component, variant, or token only when the required task cannot be expressed by the current system.
57
-
58
- Do not introduce raw brand colors, a parallel token layer, a second component system, or a speculative abstraction.
59
-
60
- ## Visual Language
61
-
62
- - Brand green is `hsl(152 44% 26%)` in light mode and `hsl(152 46% 34%)` in dark mode. It is the only brand accent.
63
- - Use the canonical warm-neutral and semantic tokens. Status and chart colors communicate meaning; they are not competing accents.
64
- - Use surface luminance steps for elevation. Keep shadows rare and subtle.
65
- - Use 6px buttons, 12px cards and surfaces, and full pills.
66
- - Use Geist Sans for interface and editorial text. Use Geist Mono for code, file names, identifiers, and timestamps.
67
- - Use Title Case for headlines and sentence case for subheads, labels, buttons, and navigation.
68
- - Keep display text tightly tracked. Hero headlines use 800 weight; section headings use 700.
69
- - Use Lucide outline icons with a 2px stroke: 16px inline, 20px in card headers, and 40px in icon tiles.
70
- - Use the existing geometric aperture mark. Do not redraw it or create a wordmark variant. The lockup is **Felan AI**; body copy says **Felan**.
71
- - On dark backgrounds, place the green mark on its established cream protection tile.
72
- - Center page content in a container up to 1400px with consistent `px-4 md:px-6` gutters. Use one column on mobile and add columns only when the content supports them.
73
-
74
- ## Voice and Claims
75
-
76
- Write with calm authority: confident, technical, and pragmatic.
77
-
78
- - Explain what changes for the team instead of announcing that something uses AI.
79
- - Address benefits with second-person **you**. Refer to **Felan** in the third person.
80
- - Use short declarative headings, concise explanations, and em dashes where natural.
81
- - Do not use exclamation marks or emoji in UI chrome.
82
- - Name agents by SDLC role: Developer agent, QA agent, SRE agent, or General Purpose agent. Do not create mascots.
83
- - Use terms such as background agent, session, delegation, automation, trigger, guardrails, approvals, audit log, team memory, SDLC, repo, PR, and rollout.
84
- - Keep QA-era terms such as coverage gap, regression, triage, flaky, and pass rate inside explicit QA examples only.
85
- - Avoid hype such as revolutionary, seamless, cutting-edge, supercharge, unlock, game-changer, magical, solution, ecosystem, synergy, empowers, transforms, 10x, AI-powered, AI-driven, and AI-first.
86
- - Do not position Felan as an orchestrator of third-party agents or as a replacement for engineers. Humans own direction, judgment, and approvals.
87
-
88
- Only make claims supported by the brief, including:
89
-
90
- - `Start with one repo`
91
- - `Works in Slack, GitHub, Linear, Jira`
92
- - `Every agent action logged`
93
-
94
- Never invent customer quotes, metrics, outcomes, prices, compliance claims, or setup times. Do not reuse Bugzy-era claims such as `5-minute setup`, `No config required`, `SOC2 ready`, `EUR 250/month`, ratios, or `auto-healed`.
95
-
96
- The legal line is `© 2026 Felan`.
97
-
98
- ## Interaction, Responsive Behavior, and Accessibility
99
-
100
- - Design mobile-first and verify narrow and wide layouts.
101
- - Preserve reading order when columns collapse or feature blocks reverse.
102
- - Prevent horizontal overflow, clipped headings, and controls that depend on hover.
103
- - Provide visible keyboard focus and accessible names for icon-only controls.
104
- - Use semantic HTML and native controls before custom interaction patterns.
105
- - Do not communicate status through color alone.
106
- - Keep motion restrained and functional. Avoid bounce and respect `prefers-reduced-motion`.
107
- - Cover relevant loading, empty, error, disabled, success, permission, and long-content states.
108
-
109
- ## Pre-Ship Check
110
-
111
- - The reader, task, surface, and primary action are clear.
112
- - Existing components and canonical tokens are reused.
113
- - Marketing is dark; product defaults light and supports dark.
114
- - There is no more than one green CTA per screen.
115
- - Colors, type, radii, icons, logo treatment, and copy follow Felan rules.
116
- - No QA-era framing, invented proof, fake screenshot, or unavailable asset remains.
117
- - Responsive layouts, keyboard focus, contrast, content states, and reduced motion were checked.
118
- - No unnecessary component, token, framework, or abstraction was introduced.