chore(deps): ceiling agt-core below 5.0 so pip cannot break Cedar - #473
Merged
Conversation
PolicyEvaluator imports agent_os.policies.backends.CedarBackend. AGT removed that module in #3444, when the v4 policy language was replaced by ACS v5, so it does not exist on any release after the 4.x line. We only work today because >=4.0 happens to resolve to 4.1.0, the last release that ships it. Nothing recorded that, and nothing prevented pip from choosing otherwise. Cedar evaluation is the enforcement path, so a resolver picking 5.x does not degrade cMCP, it stops it deciding anything correctly. >=4.1.0,<5.0 rather than ==4.1.0: the ceiling is the part that matters, and leaving the floor open means an agt-core 4.1.1 carrying the cryptography cap lift (#471, upstream #3614) is picked up without another PR. Tracked in #472. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
First step from #472.
The hazard
PolicyEvaluatorimports the v4 Cedar backend directly:AGT removed that module in microsoft/agent-governance-toolkit#3444, when the v4 policy language was replaced by ACS v5. It does not exist on
mainand will not be in any release after the 4.x line.agent-governance-toolkit-core4.1.0, tagged 2026-06-09, is the last release that ships it.We work today only because
>=4.0with no ceiling happens to resolve to 4.1.0, and 4.1.0 happens to be preferred because 5.0.0'sagentrust-trace<0.3.0conflicts with ouragentrust-trace>=0.5. Both of those are accidents of the current dependency graph, not decisions. Either could change without us touching anything.Cedar evaluation is the enforcement path, so this is not a degradation risk. A resolver that picks 5.x does not make cMCP slower or noisier, it stops it deciding anything correctly, and the demos, the web console, and every conformance claim go with it.
Why a range and not an exact pin
#472 suggested
==4.1.0.>=4.1.0,<5.0is better:<5.0is whereCedarBackenddisappears, and semver says a 4.x minor should not remove public API, so the major boundary is the correct line.CI will be red, and not because of this
The security scan fails on this branch exactly as it does on every other cmcp branch right now:
cryptography48.0.1 reports CVE-2026-69249 and CVE-2026-69248, both reachable through the untrustedcert_chaina claim carries intoverify_ak_ek_chain. That is #471, it predates this change, and this PR neither helps nor worsens it. Verified separately that withcryptography49.0.0 forced over the cap, agt-core 4.1.0 is fine: 1041 tests pass,CedarBackenddecisions are unchanged, and the audit exits clean.This PR is only about making the constraint deliberate rather than accidental.
🤖 Generated with Claude Code