This skill is a downstream specialization of the ctf-sandbox-orchestrator skill. It guides an agent through an OAuth or OIDC flow in a CTF sandbox challenge: the entry route, authorize redirect, callback, token exchange, refresh and the service that finally accepts the identity.
It is meant for challenges where the hard part is showing how a flow is shaped and accepted. The agent records scopes, state, nonce, PKCE material, redirect URIs and token claims, then reduces the result to the smallest redirect-to-acceptance sequence. The skill replies in Simplified Chinese unless English is requested.
What You Get
- A step order for mapping entry route, authorize request, callback, token exchange and refresh
- A checklist of redirect-side, token-side and acceptance-side evidence to keep together
- A rule that separates token possession from the point where a service accepts the identity
- A pointer to the JWT claim confusion skill when the problem shifts to token validation
Common Use Cases
- Explaining how a CTF login flow turns a callback code into an accepted session
- Documenting scopes, audience, subject and expiry of tokens seen in a challenge
- Reducing a long redirect trail to one canonical flow plus one minimal variant
What Makes It Unique
It is narrow: it covers only the OAuth and OIDC chain and hands off to sibling skills for JWT validation or general web routing. It insists on proving backend acceptance, not only browser-visible redirects, and it only runs after the orchestrator skill has set sandbox assumptions.
Under The Hood
The skill is instruction text only: a SKILL.md, one reference checklist and an agent config file that disables implicit invocation. It ships no scripts. It does not tell the agent to fetch pages, call APIs or write files itself, though the agent’s own tools may be used to observe the challenge flow. It requires the orchestrator skill to be active first.