Works with the AI Agents you already use
Claude Code
Cursor
Codex CLI
DeepSeek
Muse+22 moreClaude CodeCursorCodex CLIDeepSeekMuse
Dots
Manus
Grok
Grok Bot+18 agentsRepair HTTP Cache Vary Key Bugs
Repair HTTP cache key bugs around Vary that serve compressed, localized, authenticated, or tenant-specific responses to the wrong request variant.
$1
Repair HTTP Cache Vary Key Bugs
Example session with this skill installed
Repair our Node response cache: the store path includes Accept-Encoding after reading Vary, but lookup uses only URL, so identity clients sometimes receive gzip bytes. Build a fake origin and deterministic tests. Preserve TTL, eviction, revalidation, authorization, and response headers.
- Read your context and instructions
- Compiled the repair http cache
Variant-index defect found. The cache stores representations under a Vary-aware key but performs initial lookup through a URL-only alias that can point at the last stored encoding. The repair introduces one canonical variant index, normalizes only declared Vary headers, keeps TTL/eviction/revalidation policy unchanged, and verifies gzip, identity, locale, authorization, and 304 paths with a fake origin.
Connects securely to your tools. The creator never sees your data.
About this skill
The problem
HTTP caches often serve incorrect variants when lookup and store keys disagree or when the Vary header is processed inconsistently. This leads to leaking authenticated content across sessions, serving compressed bytes to legacy clients, or 304 revalidations updating the wrong representation.
What it does
- Identifies discrepancies between cache lookup and storage paths in repository code.
- Normalizes request-header selection for
Accept-Encoding,Locale, andAuthorization. - Aligns variant indexing to ensure the cache reconciles
Varyheaders learned after the origin response. - Repairs repository-level cache-key construction and tenant boundaries.
- Generates deterministic local test fixtures to verify cache hits and misses across headers.
Frameworks & tools
Works with repository-level code for CDNs, application-level cache adapters, and proxy configurations. Requires access to Read, Edit, and Bash tools.
Why this beats prompting it yourself
Generic prompts often suggest simple TTL changes or cache clearing which mask the underlying logic error. This skill enforces a strict resource register and failure timeline analysis to fix the architectural disagreement between how keys are generated and how variants are stored.
Use cases
- Fixing localized content leaks where one user sees another's language or tenant data.
- Repairing broken 304 Not Modified flows that refresh the wrong cached representation.
- Ensuring identity-encoded clients never receive Gzip or Brotli payloads.
- Standardizing cache key normalization across multi-layer CDN and app architectures.
Known limitations
Does not modify TTL, eviction policies, or origin response headers. Cannot access production environments, live secrets, or deployed state; it is limited to repository-level repairs.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Cannot choose or change TTL, eviction, cacheability, authorization, tenant isolation, revalidation, and origin response policy; cannot verify production infrastructure or prove every runtime, proxy, network, or scheduler ordering.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 2 months ago
Permissions
Run on a clean branch with repository files and disposable local fixtures. Review every patch and validation result before merge. Keep production services, real credentials, customer payloads, and deployment out of scope.