Reported issues for Brandsystem MCP — Brand Guidelines for AI
Pod holds 8 of 8 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.
Back to Brandsystem MCP — Brand Guidelines for AI.
Most discussed
Preflight: resolve same-document CSS variables and detect
logos (Colovore field run)
From the v0.13.0 Colovore field run (#41):
brand_preflightflaggedvar(--cv-font-text)literally as a "non-brand font" even though the:rootdefinition was in the same document. Preflight should resolve same-document custom properties before rule checks.- Logo-presence rule (L-PRESENT) reported "no logo detected" on a card containing
<img src="colovore-wordmark.svg" alt="Colovore">. Detection should recognize img filenames matching logo/wordmark/brand patterns and alt text matching
Read the thread · 2026-07-17 · closed · 2 comments
Robustness findings from 0.14 ingestion fuzzing: extractLogos superlinear DOM cost; raw parser errors reach agents
Two reported (not yet fixed) findings from the 0.14 malformed-input suite (test/security/ingestion-robustness.test.ts):
extractLogosis superlinear in DOM size — measured ~150ms @5k elements, ~1.1s @20k, ~7.3s @50k (cheerio.load itself <60ms; the selector + isInLogoCloud passes dominate). Bounded memory and terminates, but a hostile/huge page costs ~8s per extraction. Candidate fix: cap candidate elements scanned (e.g. first N img/svg/link matches) or early-exit once enough candidates
Read the thread · 2026-07-18 · closed · 1 comment
Fix font-family compliance parsing across newline boundaries
The 0.11 deterministic eval suite exposed a parser edge case in the compliance path.
Reproduction
A font-family declaration without a trailing semicolon can cause the current regex to consume across a newline into the next declaration or rule.
Current pattern appears in both:
src/tools/brand-check-compliance.tssrc/lib/content-scorer.ts
/font-family\\s*:\\s*([^;}"]+)/gi
Expected
Font-family extraction stops at CSS declaration, rule, or newline boundaries and pres
Read the thread · 2026-07-17 · closed · 1 comment
Request to open a Security Advisory
Hi, I have a security vulnerability to report. Please open a Draft Security Advisory so I can submit the details and PoC privately as per your security policy.
Read the thread · 2026-05-07 · open · external user · 1 comment
Same-theme color collapse: one color per (role, theme) drops legitimate same-role colors
Residual gap split out of #35 (whose four items all shipped by 0.14.x). The (role, theme) merge key preserves colors across themes, but two same-role colors within the same theme still compete for a single slot. Three labeled cases in the corpus:
- clean-semantic: body white and
--surface-muted#f4f4f4are both default-theme surfaces; white folds into the occupied slot. - messy-inline: body text
#222222loses the default-theme text slot to the more frequent#ffffff; dark bro
Read the thread · 2026-08-27 · open · 0 comments
Agent-experience papercuts from Colovore field run (unknown args, profile-unaware guidance, alpha-hex, onramp noise)
From the same v0.13.0 field run as #41, in the reporter's priority order:
- Silent unknown-argument swallowing —
{url: ...}instead ofwebsite_urlwas accepted without complaint and auto mode scaffolded an empty brand. Cost a full round-trip with zero signal. Fix: detect unknown keys centrally in safeParseParams and warn with a did-you-mean. - Core profile recommends tools it doesn't expose — brand_start's source menu, brand_status next steps, and timeout recovery all point at ful
Read the thread · 2026-07-17 · closed · 0 comments
Uncertain primary poisons the entire check loop and clarify cannot fix it (Colovore field report)
Field report from a real agent run (v0.13.0, core profile, colovore.com — a Webflow site, ground truth known from a parallel manual encoding).
The structural failure: static-CSS extraction crowned achromatic #f0f0f0 as primary (role inferred from a raw CSS variable name) while the true brand green #00A050 sat correctly identified in chromatic_candidates with suggested_primary: null. The system knew it was unsure — then compiled anyway. Downstream, every check inverts: brand_check
Read the thread · 2026-07-17 · closed · 0 comments
Adoption prompts under-route to brand_start in first-tool eval (66.7% overall)
First published eval run (0.11.0, claude-haiku-4-5-20251001, 2026-07-17 — see eval/RESULTS.md): first-tool selection scored 8/12. All four misses cluster on the adoption path:
- "How do I use my existing brand guidelines with AI?" → picked
brand_runtime(expectedbrand_start) - "I have a PDF brand guide." → picked
brand_status(expectedbrand_start) - "Use our Figma library for our brand." → picked
brand_brandcode_connect(expectedbrand_start) - "Write this LinkedIn post in our voi
Read the thread · 2026-07-17 · closed · 0 comments
Most recent
The remaining reports are on the project's issue tracker.