Competition JWT Claim Confusion is a downstream skill of the CTF Sandbox Orchestrator. It covers JWS, JWE and nested tokens, and asks how headers, claims and key selection turn into an accepted identity.
The agent splits the token path into parse, key lookup, signature or decryption, claim validation and final acceptance, and records header fields, claims and the key source before changing anything. It keeps one canonical accepted token path and, where a bypass depends on a delta, one altered path next to it.
What You Get
- A five-part split of the token path from parse to acceptance
- A header checklist and a claim checklist
- Guidance to separate token parse success from authorization success
- A rule to tie parser behavior, trust policy and resulting privilege into one chain
- A hand-off to the OAuth skill when the task grows into a redirect chain
Common Use Cases
- Explaining how a role claim becomes route access in a challenge app
- Tracing key lookup through a JWKS endpoint, local file or cache
- Checking issuer and audience matching and claim normalization
- Documenting why a decoded token was or was not accepted
What Makes It Unique
It treats the problem as a validation chain and not as a single token trick: key selection, normalization and authorization are each recorded. The skill also requires proof of acceptance by the service, since a token that decodes cleanly says nothing about whether the app grants access.
Under The Hood
The skill is instruction text only: the SKILL.md body, one reference checklist and a small agent config file that disables implicit invocation. It ships no scripts. The instructions tell the agent to inspect tokens and services in the sandbox the user supplies, to reply in Simplified Chinese unless asked for English, and to run only after the orchestrator skill is active.