Promotes the carousel label from <p> to <h2>, links the section to the
heading via aria-labelledby, adds an accessible label per chip, and
turns the opaque logo <img> tags into proper lazy-loaded assets with
explicit dimensions.
Adds a subline naming seven brands (Docker, Microsoft, HashiCorp,
Vercel, Google Cloud, AWS, Anthropic) so the homepage now has
indexable text referencing actual adopters — previously the brand
information only lived inside animated chips.
Emits a schema.org ItemList of Organization entities (JSON-LD) on the
homepage, which none of the comparable OSS sites (Vite, Biome, Astro,
Nx, Turborepo) currently expose — cheap differentiator for rich
search results.
Drops the dedicated /adopters page, its Vue component, the navbar entry,
the sidebar hack, and the contributing-guide section. The homepage
carousel already carries the social-proof signal — a separate page
attracted virtually no traffic on comparable OSS sites and added
maintenance surface without a clear payoff.
Also removes the "See all" CTA from the carousel header now that there
is nowhere to send visitors to, and centers the remaining label.
The find-adopters Go tool moves out of the repo (to ../find-adopters/)
— it was always a one-off analysis helper, not code that ships with
Task. The adopters.ts file remains the submission surface for anyone
motivated enough to PR a new entry.
Runs the refreshed find-adopters tool against every public Taskfile on
GitHub (1190 unique repos, 13 min) and uses the findings to swap in four
higher-signal entries: Azure/Azure-Sentinel replaces Microsoft's niche
Fabric provider, flet-dev/flet (16k stars, #1 non-Task hit), Anthropic's
Rust protobuf, and charmbracelet/glamour join the list. Gogs drops out
since it no longer surfaces in the best-match slice GitHub exposes.
Rewrites the discovery strategy: GitHub Code Search caps at 1000 results
per query and its size: qualifier turned out unreliable (non-monotone
total_count, sporadic 404s), so the tool now paginates each of the four
Taskfile variants to the cap and supplements with an org: scan over
~100 curated organizations. That's the practical ceiling without GH
Archive or BigQuery, and it captures every big-brand hit we care about.
Also drops the code-search rate from 24 to 8.5 req/min to match the
real 10 req/min authenticated limit.
Reshuffles the adopters list to lead with big brands (Docker, HashiCorp,
Microsoft, Vercel, Google Cloud, AWS) followed by high-profile OSS
(FerretDB, Tyk, Outline, etc.), so the section reads as strong social
proof rather than a niche catalog.
Adds website/scripts/find-adopters, a small Go CLI that sidesteps the
1000-result cap on GitHub Code Search by partitioning queries per star
bucket (and per pushed-year when a bucket overflows), then enriches
every hit via a batched GraphQL call. The result is a ranked TSV/JSON
of adopter candidates, filterable by min stars and owner type, that
can be rerun periodically to keep the list fresh. Exposed via
`task find-adopters` in website/Taskfile.yml.
Introduces a new /adopters page listing notable OSS projects using Task,
along with an infinite-scroll carousel on the homepage linking to it.
The adopter list lives in .vitepress/adopters.ts for easy PR-based
submissions; contributing docs explain the process.
This is part of the LLM plugin. It's distracting and not really useful.
We're keeping the markdown version of the pages, tho. Just append `.md`
to any page to see the markdown version.
Add //nolint:gosec annotations for intentional code patterns
that are safe in context (path traversal in release tool,
uintptr conversion for terminals, weak rand in tests,
TLS skip verify for user-configured insecure mode).
Assisted-by: Kimi-K2.5 via Crush <crush@charm.land>
