docs(auth): record the OAuth auto-approve trust model + J1 residual#119
Merged
Conversation
Release-review J1 flagged that auto-approve fires on a trusted redirect_uri alone (no client_id requirement), and open RFC 7591 registration lets any client self-register the allowlisted callback to skip consent. The adversarial-verify pass confirmed this is NOT exploitable: matchesAllowlist is exact-match and the code is 302'd only to the allowlisted URI (the real first party), the code is bound to the initiating client at /oauth/token, and the grant carries the single-tenant 'limner-dogfood' identity with the pinned ['mcp'] scope. The "obvious fix" — also requiring a trusted client_id — is intentionally NOT applied: the live agent's client_id ROTATES, so an AND with isTrustedClient would break agent auto-approve. Documented the deliberate trust decision, the non-exploitable residual, and the condition under which to revisit it (a real per-user identity replacing 'limner-dogfood'), at the auto-approve site so a future reader doesn't "harden" it into an agent-breaking change. Documentation only — no behavior, schema, or test change. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Signed-off-by: Jim Vinson <jim@vinson.org>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Final review item — J1 (documentation only). No behavior, schema, or test change.
Release-review J1 flagged that auto-approve fires on a trusted
redirect_urialone (noclient_idrequirement), and open RFC 7591 registration lets any client self-register the allowlisted callback to skip consent. The adversarial-verify pass confirmed this is NOT exploitable:matchesAllowlistis exact-match and the code is 302'd only to the allowlisted URI (the real first party), the code is bound to the initiating client at/oauth/token, and the grant carries the single-tenantlimner-dogfoodidentity with the pinned['mcp']scope.The "obvious fix" — also requiring a trusted
client_id— is intentionally not applied: the live agent'sclient_idrotates, so anANDwithisTrustedClientwould break agent auto-approve (violating the "live agent untouched" guardrail). This PR documents the deliberate trust decision, the non-exploitable residual, and the condition under which to revisit it (a real per-user identity replacinglimner-dogfood) — at the auto-approve site (consent.ts) so a future reader doesn't "harden" it into an agent-breaking change.Typecheck + lint clean; 244 mcp tests unchanged.
🤖 Generated with Claude Code