Skip to content

feat/agentcore 1 governance - #6771

Closed
kyleseaman wants to merge 2 commits into
mainfrom
feat/agentcore-1-governance
Closed

kyleseaman wants to merge 2 commits into
mainfrom
feat/agentcore-1-governance

Conversation

@kyleseaman

Copy link
Copy Markdown
Contributor

Problem / Motivation

Why it matters

What changed (motivation → approach → change)

Tests

Manual verification

Screenshots / video

Related Issues

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

Land the inert AgentIdentityProvider seam, Default no-op, and
capabilities.agentcore catalog row so later stack PRs can attach AWS,
Gateway inject, login, and UI without widening this contract. Policy
validators live in agentcore_schema so parse does not import AWS.
CapabilityGate.from_dict now rejects a non-boolean enabled the
same way ScopedRuleset.mode does, so enabled: "false" cannot
permit any capability row. Drop the agentcore-only branches.
@kyleseaman
kyleseaman requested review from a team and krishdhasmana August 29, 2026 11:39
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Aug 29, 2026
@kyleseaman

Copy link
Copy Markdown
Contributor Author

Opened by accident while trying to rename the AgentCore stack off the cursor/ prefix. The live stack remains #6594–#6602. Closing this duplicate.

@kyleseaman kyleseaman closed this Aug 29, 2026
@github-actions github-actions Bot removed the readiness: checking Automated validation is still running label Aug 29, 2026
@github-actions

Copy link
Copy Markdown
Contributor

Design Review (Fable 5) — 🟡 CONCERNS

Design-level review of bd8eab069b781429f41b52cc508182cae2474f0a — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

Sound, staged seam that follows the platform's own precedents — but the PR description is a blank template on 2.6k lines of new public surface.

Watch

  • Description ↔ diff fidelity is zero. The body is the unfilled template — no problem statement, no test/verification sections — while the diff adds a new PlatformContext slot, a Protocol on the v1 contract, three policy-only ceiling fields, a repo-wide CapabilityGate.enabled strictness change, and a reserved secret directory. The RFC/plan docs carry the why, but the PR body is what reviewers and git log archaeology see; fill it before merge, explicitly flagging the behavior change that a policy with a coerced enabled (e.g. "false", which previously enabled the gate) now refuses to parse.
  • Seam lands ahead of its consumer. _agent_identity_enabled in agent.py has only test callers; that matches the repo's reserved-slot pattern and the plan doc names the follow-up, so acceptable — but it stays dead weight if the series stalls.

Suggestions

  • _capability_raw_for_gate and the three _apply_agentcore_* passes hardcode one scope name into parse_policy; a policy_only_fields entry on ScopeSpec would keep the parser scope-agnostic and stop the next payload-carrying capability from copying the branch.

[DESIGN-REVIEWED] bd8eab0

@github-actions

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5) — 🟡 CONCERNS

Premise-level review of bd8eab069b781429f41b52cc508182cae2474f0a — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

All evidence gathered. Counts verified: _agent_identity_enabled has zero consumers in src/ (grep _agent_identity_enabled — definition plus tests only); the composed agentcore_gateway_url/agentcore_workload_name ceiling fields and their public readers have zero src/ consumers; six of the seven AgentIdentityProvider methods are called nowhere; and the plan doc commits Task 5/6 steps as [x] while none of that code exists (glob test/test_agentcore_*.py → only test_agentcore_schema.py). Here is the review.

First-Principles-Verdict: CONCERNS

The PR description is an unfilled template, so every item is undeclared — and two surfaces (policy gateway_url/workload_name, the agent.py probe) land with zero consumers.

What this change ships

Intent: land the design record plus the dormant governance/CPP seams for AgentCore agent identity, ahead of the code that will use them — an ADDITION (first PR of a declared stack).

  1. AgentCore identity/gateway RFC + README row recorded — justified
  2. Implementation plan doc — justified, but Task 5/6 steps pre-checked [x] for unshipped code
  3. Policies can enable opt-in capabilities.agentcore; enable-without-known-posture fails closed — justified, undeclared
  4. Policies can carry gateway_url/workload_name, validated at parse — zero consumers
  5. Non-boolean enabled on ANY capability now aborts instead of coercing to on — rides along, undeclared, cause-level fix
  6. New agent_identity platform slot + protocol + no-op default — 6 of 7 methods zero consumers
  7. agent.py _agent_identity_enabled three-conjunct probe — zero consumers
  8. agentcore-inbound added to the secret deny list before its writer — justified (fail-closed ordering)
  9. Profiles carrying posture/URL/name keys rejected at parse — justified (policy-only trust root)
  10. Spec/constants updates in the same commit — justified (documented invariant)

Watch

  • Description is the untouched template; the intent lives only in the in-diff RFC. Item 5 is the one a human must see: bool("false") previously left a default-ON capability (memory_writes, browse, …) enabled, and the fix now aborts boot on any policy that was silently mis-parsing — right direction, but a behavior change to shipped policies that nothing declares.
  • Zero-consumer staging is the plan-of-record (knowledge/dashboard/jail precedent, pre-launch, CONTRACT_VERSION unbumped), so it clears the one-way-door bar — but gateway_url/workload_name belong to plan Task 7/2b, not Phase 1, and nothing composed onto the ceiling is read (grep agentcore_gateway_url|agentcore_workload_name in src/: parse-side writes only, 0 readers).

Subtractions

  • Defer GovernanceCeiling.agentcore_gateway_url / agentcore_workload_name, their _apply_* parsers, and the two public readers (governance.py:1975, :2015) to the PR that consumes them — 0 consumers counted; keep only posture, which the fail-closed rule needs.
  • Defer _agent_identity_enabled / _agentcore_capability_permitted (agent.py:97-146) to the first chokepoint PR — 0 callers in src/.
  • Remove the [x] marks on plan Tasks 5 and 6 (docs/superpowers/plans/2026-08-27-agentcore-identity-gateway.md:432-489) — the plan instructs agentic workers to track by checkbox, and none of that code is in this repo.

[FIRST-PRINCIPLES-REVIEWED] bd8eab0

@github-actions

Copy link
Copy Markdown
Contributor

Opus 4.8 Review — ✅ no blocking findings

Reviewed bd8eab069b781429f41b52cc508182cae2474f0a — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] bd8eab0

Verdict parsed from the review's SHA-scoped output markers for commit bd8eab069b781429f41b52cc508182cae2474f0a.

False positive or not applicable? A repository writer can comment:
/ai-review override fable bd8eab069b781429f41b52cc508182cae2474f0a: <one-sentence reason>

@github-actions

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of bd8eab069b781429f41b52cc508182cae2474f0a and found no blocking issues.

This comment is updated in place on each push.

Review details

FINDING -- src/kiro_crew/platform/governance.py:1948 -- Function-local validator imports at lines 1948 and 1988 violate top-level-imports -> Fix: move both imports to the module import block.
[GPT-REVIEWED] bd8eab0

False positive or not applicable? A repository writer can comment:
/ai-review override gpt bd8eab069b781429f41b52cc508182cae2474f0a: <one-sentence reason>

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant