Repository navigation
Per-agent keys bound to mandates, and a threat model - #2
Merged
Merged
Conversation
Until now every agent shared one API key and named itself in the request body, so any agent could act as any other and use any mandate. - charter agent add / list / rotate / revoke. Each agent gets its own key, bound to specific mandates. The key is shown once; only its SHA-256 hash is recorded. Registry state is derived from the audit log. - The agent's identity comes from its key, never the request body. A body naming a different agent is refused (403), as is a mandate the agent is not bound to. - An agent reads only its own status and escalation results. Someone else's is reported as not found, so its existence is not revealed. - Rotation replaces the key immediately. Revocation is immediate. - The shared CHARTER_API_KEY still works; CHARTER_REQUIRE_AGENT_KEYS=true refuses it and accepts only registered agents. - All credential comparisons are now constant-time. A live run against the real server found a flaw the tests had not: revoking the last agent dropped the server back to open mode, where anyone could call it. Once any agent has been registered the API now stays closed. Regression tests added at the service and API level. docs/THREAT_MODEL.md lists what each control defends against and, as plainly, what it does not. README trust-boundary section updated. 27 new tests (122 total).
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Every agent used to share one API key and name itself in the request body, so any agent could act as any other and use any mandate.
Each agent now gets its own key bound to specific mandates, its identity comes from the key rather than the body, it can only read its own results, and keys can be rotated and revoked immediately. Only a hash is stored. The shared key still works unless CHARTER_REQUIRE_AGENT_KEYS=true.
A live run found a flaw the tests had missed: revoking the last agent dropped the server to open mode. Once any agent has been registered the API now stays closed, with tests at the service and API level.
docs/THREAT_MODEL.md lists what each control defends against and what it does not.
27 new tests (122 total). Verified live: identity from key, spoofing and wrong-mandate refused, cross-agent reads hidden, rotation and revocation immediate, and the full serve plus rogue-agent demo still works.