- Reframe 'When NOT to use' around version-dependent correctness
- Add guidance for conflicting official sources
- Prefer deep links with anchors in citation examples and rules
Grounds every framework-specific implementation decision in official
documentation — verify, cite sources, and surface what's unverified.
Covers the full process from stack detection to citation format,
with a framework-agnostic design and optional documentation cache.
- Add width/height to <source> elements for correct aspect-ratio in art direction.
- Use realistic 'sizes' attributes in hero image examples.
- Remove 'decoding=sync' from hero images to avoid blocking the main thread.
- Align TTFB 'good' threshold with web.dev standards (800ms).
- Distinguish synthetic (Lighthouse) vs RUM (web-vitals) measurement approaches,
clarifying when each is appropriate: synthetic for CI regression detection,
RUM to validate real user impact.
- Expand TTFB diagnosis from a single vague hint into a decision tree that breaks
down each component (DNS, TCP/TLS, server processing) with specific next steps.
Mirrors the tree in the skill and adds a dedicated checklist section.
- Fix image optimization example: the previous "GOOD" example applied loading="lazy"
without distinguishing the LCP hero image from below-the-fold images. Hero images
must never be lazy-loaded. New example separates both cases explicitly.
- Add art direction + resolution switching to the hero image example using <picture>
with media queries for mobile/desktop crops and srcset for density variants.
Mobile-first: <img src> fallback points to the mobile version.
Covers AVIF → WebP → JPG format cascade and fetchpriority="high" for LCP.
- Correct the date-fns tree-shaking example: modern bundlers (Vite, webpack 5+)
handle named imports automatically. The "BAD" pattern was not actually bad,
and following it could lead to unnecessary micro-optimizations. Real gains
come from dynamic imports and route-level code splitting, which the example
now illustrates instead.
Moved test-driven-development from Verify to Build in both CLAUDE.md and README.md (table + directory tree). This reflects the feedback that tests should ship with each commit as part of the build phase, not as a separate verification step after the fact.
Runs on every push and PR:
1. Validates marketplace and plugin manifests
2. Tests marketplace add + plugin install end-to-end
No ANTHROPIC_API_KEY required — these are CLI/filesystem
operations, not LLM calls.
The `/install github:addyosmani/agent-skills` syntax throws
"Args from unknown skill" errors. Revert to the previous
two-step plugin marketplace commands that work correctly.
Fixes#11
Update all SKILL.md descriptions to lead with what the skill does
(third person) followed by trigger conditions, per Anthropic's
skill authoring docs. All original trigger conditions preserved.
Update description max chars to match Anthropic spec (1024).
Retain workflow-summary warning in skill-anatomy.md to prevent
agents from following descriptions instead of reading full skills.
Add Table of Contents to reference files longer than 100 lines so
Claude can see the full scope of available information even when
previewing with partial reads.
Anthropic's Claude 4 prompting best practices document warns that
newer models overtrigger on strong directive language (Never/Always/MUST)
that was necessary for older models. This replaces aggressive directives
with natural phrasing across skills and agents while preserving all
security-critical rules (OWASP, prompt injection, secrets) unchanged.
Also reframes trunk-based development as "Recommended" with a note
acknowledging alternative branching strategies like gitflow.
Provides a structured approach for evaluating audit findings by
severity, reachability, and fix availability. Helps distinguish
between findings that need immediate action versus those that can
be tracked and fixed on a regular schedule.
Added a decision tree and example for speeding up CI: dependency
caching, parallel jobs, path filters, matrix sharding, test suite
optimization, and larger runners. Ordered by typical impact.
Replaced vague description with concrete numbers: agent focus degrades
above ~5,000 lines of non-task-specific context. Recommended target is
under 2,000 lines of focused context per task.
The table previously listed AI defaults and their alternatives but did
not explain why each pattern is problematic. Added a middle column with
the specific design or UX reason to avoid each default.
Added concrete signals for when a task needs further breakdown: session
length, acceptance criteria count, subsystem scope, and the "and" test
for task titles that indicate multiple tasks in one.
Added a concrete table with green/yellow/red thresholds for error rate,
P95 latency, client JS errors, and business metrics. Gives clear
criteria for advancing, holding, or rolling back at each rollout stage.
Added a decision tree for when bugs cannot be reproduced on demand,
covering timing-dependent, environment-dependent, state-dependent,
and truly random failure modes with specific investigation steps.
Helps determine what to measure first based on the symptom: slow first
load, sluggish interactions, navigation delays, or backend issues. Each
branch points to a specific profiling action.
The implementation phase now points to context-engineering for guidance
on loading the right spec sections per task instead of dumping the full
spec into context.
