feat(collab): agent delegation handshake + sponsor_contact_id + cascades (D1) - #2048
feat(collab): agent delegation handshake + sponsor_contact_id + cascades (D1)#2048hognek wants to merge 22 commits into
Conversation
Qodo reviews are paused for this user.Troubleshooting steps vary by plan Learn more → On a Teams plan? Using GitHub Enterprise Server, GitLab Self-Managed, or Bitbucket Data Center? |
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (12)
Included review availability: Your plan provides up to 2 included reviews per hour; 0 remain after this review. 📝 WalkthroughWalkthroughAdds cross-user delegation with scope validation, sponsor-aware registration, invite metadata, automatic or manual approval, PIN-free redemption, cascade revocation, project settings, and peer kill switches. ChangesDelegation and persistence
Estimated code review effort: 5 (Critical) | ~90 minutes Merge Risk: 🟠 High · up to This change adds cross-user delegation, persistent sponsored identities, approval, and revocation controls, but the current implementation can overgrant scopes, leave delegated access outside revocation, strand approved requests, or report a contact pause while access remains active. These security and correctness issues make the PR unsafe to merge until fixed or explicitly accepted by the appropriate owner. Suggested reviewers: 🚥 Pre-merge checks | ✅ 3 | ❌ 2❌ Failed checks (2 warnings)
✅ Passed checks (3 passed)
Full details: Linked Issues checkExplanation The PR implements the main requirements in [ Resolution Ensure complete_delegation_approval delivers the minted invite and approval result to the original requester through the peer channel. Add an end-to-end test for the manual approval delivery path. Also surface sponsor reassignment failures instead of silently ignoring them if sponsor assignment protection is required for this issue scope. Full details: Out of Scope Changes checkExplanation The changes are aligned with [ Full details: Docstring CoverageExplanation Docstring coverage is 41.71% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 199 functions across 29 files. (1 skipped: 1 unsupported.) ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
|
![kilo-code-bot[bot]](https://avatars.githubusercontent.com/in/2193792?s=60&v=4){kind=small}
| reply = "delegation denied" | ||
| elif completed: | ||
| reply = "delegation approved - the task has been assigned" | ||
| else: |
There was a problem hiding this comment.
CRITICAL: reply is unbound on the approved path — regression from this PR's edit.
The original code had if not approved: ... elif completed: ... else: ... but this change deleted the if not approved: and elif completed: conditions, collapsing them into a bare else:. Now reply = f"The delegation to {to_agent} was denied." is only assigned in the denied branch, while await _route_answer_to_agent(decision, reply) at line 420 always runs. On any approved delegation this raises UnboundLocalError/NameError, breaking the existing #161 gated-delegation approval flow (not just the new collab gate).
| else: | |
| if not approved: | |
| reply = "delegation denied" | |
| elif completed: | |
| reply = "delegation approved - the task has been assigned" | |
| else: | |
| reply = "delegation approved, but assigning the task failed - please retry" |
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| return {"status": "error", "error": "registry not available"} | ||
|
|
||
| # Find all active agents sponsored by this contact. | ||
| sponsored = await registry.list_by_sponsor(contact_id, status="active") |
There was a problem hiding this comment.
WARNING: project_id is documented as restricting revocation to a single project, but the actual agent revocation ignores it.
list_by_sponsor(contact_id, status="active") returns every sponsored agent for the contact regardless of project, so set_status(..., "revoked") revokes ALL of the contact's sponsored identities across every project. Only the later task-unassign step honors project_id. The docstring says "When project_id is provided, only revoke tokens bound to that project (membership-revoke)" — the behavior contradicts the contract and causes over-revocation. Filter the sponsorees by project when project_id is set.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| contact_id = decision_metadata.get("contact_id", "") | ||
| agent_slug = decision_metadata.get("agent_slug", "") | ||
| display_name = decision_metadata.get("display_name", "") | ||
| granted_scopes = decision_metadata.get("granted_scopes", []) |
There was a problem hiding this comment.
WARNING: complete_delegation_approval mints the invite directly from stored granted_scopes without re-applying validate_delegation_scopes or binding to the authorized project.
The stored granted_scopes come from the original decision metadata. If those ever contain files_write/decisions_write they are minted verbatim — the hard-denylist is only enforced at request time (process_delegation_request), not at approval-completion time. Re-run validate_delegation_scopes(...) here and refuse/ strip any deny-scoped invite, and assert project_id matches the project the decision was created for, so a tampered or legacy decision can't mint an over-privileged invite.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| Reversible by clearing the flag. | ||
| """ | ||
| # Set a flag that the peer routes check. | ||
| request.app.state._peer_disabled = True |
There was a problem hiding this comment.
SUGGESTION: The per-instance panic flag has no re-enable path.
kill_switch_per_instance sets request.app.state._peer_disabled = True and the docstring claims "Reversible by clearing the flag", but no function clears it (and nothing resets it on restart). Operators have no supported way to resume peer traffic after a panic. Add an explicit resume/clear function or document that a restart is required.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| try: | ||
| from tinyagentos.projects.a2a import ensure_a2a_channel | ||
|
|
||
| if project_id: |
There was a problem hiding this comment.
SUGGESTION: A2A audit line is only emitted when project_id is set.
The "Collaboration with ... has been revoked" system message is guarded by if project_id:, so a contact-wide revoke (project_id=None) leaves no in-channel audit trail of which agents/tasks were revoked. Consider posting the summary to a default/admin channel or logging it even when project_id is absent.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
Code Review SummaryStatus: 4 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)SUGGESTION
Files Reviewed (5 files, 3 new commits since 7be40ca)
Fix these issues in Kilo Cloud Previous Review Summaries (13 snapshots, latest commit 7be40ca)Current summary above is authoritative. Previous snapshots are kept for context only. Previous review (commit 7be40ca)Status: 5 Issues Found | Recommendation: Address before merge Overview
This is an incremental review against the four new commits since the previous review at Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (4 files)
Fix these issues in Kilo Cloud Previous review (commit f0bc711)Status: 4 Issues Found | Recommendation: Address before merge Overview
This is an incremental review against the two new commits since the previous review at Issue Details (click to expand)WARNING
SUGGESTION
Files Reviewed (7 files)
Fix these issues in Kilo Cloud Previous review (commit bb0cd89)Status: No Issues Found | Recommendation: Merge Files Reviewed (6 files)
Previous review (commit 46a3fbf)Status: No Issues Found | Recommendation: Merge Overview
Files Reviewed (6 files)
Previous review (commit 2d52981)Status: No Issues Found | Recommendation: Merge Overview
Files Reviewed (6 files)
Previous review (commit 053581e)Status: 8 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)CRITICAL
WARNING
Files Reviewed (16 files)
Fix these issues in Kilo Cloud Previous review (commit eb604e5)Status: 4 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)CRITICAL
WARNING
Files Reviewed (2 files)
Fix these issues in Kilo Cloud Previous review (commit f2e727e)Status: 4 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
Files Reviewed (2 files)
Fix these issues in Kilo Cloud Previous review (commit 5503bac)Status: No Issues Found | Recommendation: Merge Files Reviewed (3 files)
Previous review (commit d0f6628)Status: No Issues Found | Recommendation: Merge Files Reviewed (2 files)
Previously Reported Issues — Now ResolvedAll 3 outstanding findings from the prior incremental review (at
Previous review (commit 68baed6)Status: No Issues Found | Recommendation: Merge Files Reviewed (2 files)
Previously Reported Issues — Now ResolvedAll 3 outstanding findings from the prior incremental review (at
Previous review (commit cccdaf9)Status: 3 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)WARNING
SUGGESTION
Previously Reported Issues — Now ResolvedThe 5 findings from the prior review (at
Files Reviewed (2 changed since prior review)
Fix these issues in Kilo Cloud Previous review (commit dc2caa1)Status: 5 Issues Found | Recommendation: Address before merge Overview
Issue Details (click to expand)CRITICAL
WARNING
SUGGESTION
Files Reviewed (9 files)
Reviewed by minimax-m3:free · Input: 48.6K · Output: 7.9K · Cached: 575.6K |
| # approval. The denylist is the authoritative gate; the stored scopes are | ||
| # the human-approved set, but the code must never mint tokens for | ||
| # hard-denied scopes regardless. | ||
| safe_scopes, re_denied = validate_delegation_scopes(granted_scopes) |
There was a problem hiding this comment.
WARNING: Scope denylist is re-applied here (good), but project-membership is still NOT re-validated at approval time. This is the remaining half of the earlier complete_delegation_approval finding (#2): process_delegation_request checked is_project_member(contact_id, ..., member_kind="human") at request time, but a Decisions card can be approved much later. If the contact was removed from project_id (or the decision metadata was tampered) between request and approval, this still mints a live project invite for a project the sponsor no longer belongs to. Re-run is_project_member(project_id, contact_id, member_kind="human") here and refuse when it fails.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| for proj in (memberships or []): | ||
| pid = proj.get("id") or proj.get("project_id") | ||
| if not pid: | ||
| continue |
There was a problem hiding this comment.
WARNING: list_projects_for_member is called only inside a bare except AttributeError: memberships = []. This method is not defined anywhere in this PR (the new project_store helpers are is_project_member, get/set_project_setting), so the except AttributeError will trigger on every contact-wide revoke and silently produce no audit messages at all — defeating the entire point of this new branch. Either define list_projects_for_member on ProjectStore or drop the silent fallback so the missing-method bug surfaces loudly instead of hiding a missing audit trail for a security-relevant revocation.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| except Exception: | ||
| logger.warning( | ||
| "collab delegation completion failed for decision %s", | ||
| decision.get("id"), exc_info=True, |
There was a problem hiding this comment.
SUGGESTION: When complete_delegation_approval raises, this except Exception logs but routes no answer back to the agent. The decision answer was already persisted as approve, so the remote contact/agent is left permanently hanging with no notification that sponsor minting failed. Route a failure message (like the else branch at line 1316) inside the except so the caller learns the approval did not complete.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
|
@hognek Rebase needed - this has conflicts against current dev (GitHub reports Everything else on it looks fine from my side; it is purely staleness. Rebase onto current dev and I will take another look. |
aac8157 to
d0f6628
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 5
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tinyagentos/routes/peer.py (1)
258-272: 📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick winThe
delegation_requestdispatch sits after the "unrecognised kind" log.Every delegation request emits
unrecognised kind — accepted, no dispatchbefore being correctly dispatched, which will mislead anyone debugging the D1 flow. Move the branch above the log (and drop the redundantbody_datare-read; it's already bound at line 250).♻️ Proposed fix
+ # Dispatch delegation requests to the cross-user collab handler (D1). + if kind == "delegation_request": + from tinyagentos.delegation_handler import process_delegation_request + + result = await process_delegation_request( + request, contact_id=contact_id, envelope_body=body_data, + ) + return {"status": result.get("status", "unknown"), "detail": result} + # Log unrecognised kinds for debugging; they are accepted but not dispatched. logger.info( "peer_inbox: contact=%s kind=%s nonce=%s (unrecognised kind — accepted, no dispatch)", contact_id, kind, envelope.get("nonce", "?"), ) - # Dispatch delegation requests to the cross-user collab handler (D1). - if kind == "delegation_request": - from tinyagentos.delegation_handler import process_delegation_request - - body_data = envelope.get("body", {}) - result = await process_delegation_request( - request, contact_id=contact_id, envelope_body=body_data, - ) - return {"status": result.get("status", "unknown"), "detail": result} - return {"status": "received", "kind": kind, "nonce": envelope.get("nonce")}🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tinyagentos/routes/peer.py` around lines 258 - 272, Move the delegation_request branch in the peer inbox handler above the unrecognised-kind logger so valid delegation requests are not logged as unrecognised. Reuse the existing body_data binding instead of rereading envelope.get("body", {}), while preserving the process_delegation_request call and response mapping.
🧹 Nitpick comments (5)
tinyagentos/routes/decisions.py (1)
476-479: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueDrop the extraneous
fprefix (Ruff F541).🧹 Proposed fix
- error_msg = ( - f"Delegation approval failed: an internal error occurred while " - f"setting up the sponsored agent. Please retry or check the logs." - ) + error_msg = ( + "Delegation approval failed: an internal error occurred while " + "setting up the sponsored agent. Please retry or check the logs." + )🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tinyagentos/routes/decisions.py` around lines 476 - 479, Remove the unnecessary f-string prefix from the `error_msg` assignment in the delegation approval error path, since the message contains no interpolated expressions and Ruff F541 flags it.Source: Linters/SAST tools
tinyagentos/projects/project_store.py (1)
377-396: 🗄️ Data Integrity & Integration | 🔵 Trivial | 💤 Low valueRead-modify-write on the
settingsJSON blob can lose a concurrent key update.Two concurrent
set_project_settingcalls for different keys both read the old blob and the later write wins. Acceptable for a low-traffic knob, but worth a single UPDATE using SQLite'sjson_setif settings become hotter.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tinyagentos/projects/project_store.py` around lines 377 - 396, Update set_project_setting to perform the settings mutation in a single SQLite UPDATE using json_set, rather than reading and rewriting the entire JSON blob in Python. Preserve existing settings and support updating the requested key without allowing concurrent calls for different keys to overwrite one another; retain the method’s boolean success contract.tinyagentos/routes/project_invites.py (1)
1124-1131: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low valueDuplicated metadata-parsing block — extract a helper.
The nine-line extraction is identical in both redeem paths (and
_row_to_dictalready deserialisesmetadata, so this is purely defensive). A small_sponsor_from_invite(invite)helper keeps the two paths from drifting.Also applies to: 1224-1231
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tinyagentos/routes/project_invites.py` around lines 1124 - 1131, Extract the duplicated sponsor_contact_id metadata parsing from both redeem paths into a shared _sponsor_from_invite(invite) helper. Preserve the existing defensive handling for missing, string, invalid, and non-dict metadata, then replace both inline blocks with calls to the helper.tests/test_collab_d1_delegation.py (1)
126-300: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winAdd coverage for the delegation handler paths, not just validators and stores.
All 15 tests exercise pure helpers (
validate_delegation_scopes,_validate_delegation_envelope_body) and store CRUD.process_delegation_request,complete_delegation_approval, andcascade_sponsor_revokeare untested — a fakeapp.statewith the invite/decision/registry stores would have caught theinvite["id"]return-shape mismatch flagged intinyagentos/delegation_handler.py.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests/test_collab_d1_delegation.py` around lines 126 - 300, Add async tests covering the delegation handler functions process_delegation_request, complete_delegation_approval, and cascade_sponsor_revoke, using a fake app.state configured with invite, decision, and registry stores. Exercise successful and relevant failure/revocation paths, and assert returned values match the stores’ actual invite identifier shape, including catching any invite["id"] versus invite["invite_id"] mismatch.tinyagentos/delegation_handler.py (1)
518-522: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick winSwallowed unassign failures are invisible.
except Exception: passhides per-task failures, socountunder-reports and a revoked agent can stay assigned with no log trace. Log at warning level (Ruff S110).🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tinyagentos/delegation_handler.py` around lines 518 - 522, Update the exception handler around task_store.update_task in the unassignment flow to log each failure at warning level, including the task identifier and exception details, instead of silently passing. Preserve the existing count behavior so count increments only after a successful update, and satisfy Ruff S110.Source: Linters/SAST tools
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tinyagentos/delegation_handler.py`:
- Around line 294-298: The two mint call sites incorrectly read the invite
identifier from the top-level result. In tinyagentos/delegation_handler.py lines
294-298 and 378-382, update _auto_approve_delegation and
complete_delegation_approval to return the identifier from
invite["record"]["invite_id"] while preserving the existing approved response
shape.
- Around line 272-286: Update the invite creation calls in the current flow and
complete_delegation_approval to pass pin_required=False as the mint argument,
rather than only storing it in metadata. Keep the existing metadata and other
mint parameters unchanged so sponsored-agent invites persist a disabled PIN
requirement.
- Around line 414-420: Update the project-scoped revoke membership check in the
delegation revoke flow so a non-None project_id with a missing project_store
fails closed instead of skipping filtering. Return an appropriate error or
revoke no identities in that case; preserve the existing membership check for
available project stores and contact-wide behavior when project_id is None.
In `@tinyagentos/projects/project_store.py`:
- Around line 363-375: Update get_project_setting to validate that the project's
settings value is a dictionary before calling .get; return default for lists,
strings, or any other non-dict JSON value, while preserving the existing
behavior for missing settings and valid dictionaries.
In `@tinyagentos/routes/agent_auth_requests.py`:
- Around line 455-458: Update the existing-identity reuse flow around
registry.set_sponsor so sponsorship is only assigned when the identity’s current
sponsor_contact_id is empty; preserve any existing sponsor, including when it
belongs to another contact, and avoid re-parenting it during sponsored
redemption.
---
Outside diff comments:
In `@tinyagentos/routes/peer.py`:
- Around line 258-272: Move the delegation_request branch in the peer inbox
handler above the unrecognised-kind logger so valid delegation requests are not
logged as unrecognised. Reuse the existing body_data binding instead of
rereading envelope.get("body", {}), while preserving the
process_delegation_request call and response mapping.
---
Nitpick comments:
In `@tests/test_collab_d1_delegation.py`:
- Around line 126-300: Add async tests covering the delegation handler functions
process_delegation_request, complete_delegation_approval, and
cascade_sponsor_revoke, using a fake app.state configured with invite, decision,
and registry stores. Exercise successful and relevant failure/revocation paths,
and assert returned values match the stores’ actual invite identifier shape,
including catching any invite["id"] versus invite["invite_id"] mismatch.
In `@tinyagentos/delegation_handler.py`:
- Around line 518-522: Update the exception handler around
task_store.update_task in the unassignment flow to log each failure at warning
level, including the task identifier and exception details, instead of silently
passing. Preserve the existing count behavior so count increments only after a
successful update, and satisfy Ruff S110.
In `@tinyagentos/projects/project_store.py`:
- Around line 377-396: Update set_project_setting to perform the settings
mutation in a single SQLite UPDATE using json_set, rather than reading and
rewriting the entire JSON blob in Python. Preserve existing settings and support
updating the requested key without allowing concurrent calls for different keys
to overwrite one another; retain the method’s boolean success contract.
In `@tinyagentos/routes/decisions.py`:
- Around line 476-479: Remove the unnecessary f-string prefix from the
`error_msg` assignment in the delegation approval error path, since the message
contains no interpolated expressions and Ruff F541 flags it.
In `@tinyagentos/routes/project_invites.py`:
- Around line 1124-1131: Extract the duplicated sponsor_contact_id metadata
parsing from both redeem paths into a shared _sponsor_from_invite(invite)
helper. Preserve the existing defensive handling for missing, string, invalid,
and non-dict metadata, then replace both inline blocks with calls to the helper.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: f72dc1f2-8989-4e2b-b7d5-b3ea72fe40c2
📒 Files selected for processing (9)
tests/test_collab_d1_delegation.pytinyagentos/agent_registry_store.pytinyagentos/delegation_handler.pytinyagentos/projects/invite_store.pytinyagentos/projects/project_store.pytinyagentos/routes/agent_auth_requests.pytinyagentos/routes/decisions.pytinyagentos/routes/peer.pytinyagentos/routes/project_invites.py
|
Deep review done against head Worth saying first: roughly 6 of 8 kilo findings are stale against head, fixed by A1. CRITICAL: A2. CRITICAL: the minted invite can never be redeemed, because its PIN is generated and discarded. A3. HIGH: the approval result reaches nobody. B1. HIGH: the sponsored scope tier is not enforced. This contradicts the design's own threat model at Exploit: a contact sends Fix: intersect with an allowlist rather than subtracting a denylist, and render anything outside the tier as distinct elevated scopes needing their own approval. B2. HIGH: revocation does not cut off delegated authority, because nothing calls the cascade. Exploit: after the owner removes a contact from the project and revokes it, the sponsored agent keeps its active registry row and JWT and continues reading board, canvas and A2A indefinitely. Fix: call C. HIGH: the task-unassign cascade is a silent no-op. E. Tests run but cover none of the risk. No skip guards, all imports resolve, 15 passed and 184 across the related suites with no regressions. But coverage is confined to pure functions and store CRUD: zero tests for Things I checked that are genuinely fine: peer-inbox auth is solid (contact_id derived from the hashed bearer token, envelope from/to binding, Ed25519 signature against the pinned pubkey, nonce replay rejection, rate limit, size cap). Cross-user targeting is gated by Two unrelated nits: the Docs: no CHANGELOG entry, and Minimum bar to flip this to fix-forward: items A1, A2 and the end-to-end test are non-negotiable, because without them this is a feature that has never once run. Then A3 delivery, B1 allowlist, B2 cascade call site, and either fix or remove |
CodeRabbit findings resolved (2026-07-27 scan)All 5 findings addressed in 5503bac: 🔴 CRITICAL: Mint return shape (
|
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@tests/test_collab_d1_delegation.py`:
- Around line 376-384: Update cascade_sponsor_revoke to validate project_store
availability before calling list_by_sponsor or entering the sponsored-agent loop
whenever project_id is set. Add coverage for an empty sponsored-agent list,
asserting it returns an error containing “project store not available” rather
than “revoked,” while preserving existing behavior for available stores.
- Around line 434-437: The test currently checks the guard locally before
calling AgentRegistryStore.set_sponsor(), so it does not exercise the
implementation enforcing sponsor binding. Update the test to use
assign_approved_auth_agent_request() through the approval/reuse path for
delegated-agent sponsorship, or move the no-reparenting enforcement into
set_sponsor() and update direct coverage accordingly; verify an already
sponsored agent cannot be reassigned.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 71ff443f-a859-4919-962c-7848330e838a
📒 Files selected for processing (4)
tests/test_collab_d1_delegation.pytinyagentos/delegation_handler.pytinyagentos/projects/project_store.pytinyagentos/routes/agent_auth_requests.py
🚧 Files skipped from review as they are similar to previous changes (3)
- tinyagentos/routes/agent_auth_requests.py
- tinyagentos/projects/project_store.py
- tinyagentos/delegation_handler.py
|
Re-reviewed at head Two corrections to my own earlier review first, because I was wrong on both and you should not spend time on them. I stopped inferring and ran it. I drove So A1 is genuinely dead, no KeyError. A2 and B1 are alive and now demonstrated rather than argued: the minted invite still cannot be redeemed by anyone, and
On A2 being worse: adding On C, the reason it silently does nothing, so it can be fixed properly: N1, and this is the one I would fix first after A2: three of the five new regression tests cannot fail. I checked out N2: the policy knob cannot be turned on. N3: the new fail-closed guard is inside the loop ( On the green checks. 694 tests pass across routes, registry, contacts, decisions and projects, plus 263 at store level, and the D1 file is 20/20. No Minimum to flip this to mergeable: fix A2 so If the cascades and kill switch are actually D2 scope, please drop them from the title and file them separately rather than shipping unreferenced code. Unreferenced code that reads as a working safety control is worse than no code, because the next person assumes the kill switch works. |
A2 fix (jaylfc#2048): redeem() always validated the PIN hash regardless of pin_required. When pin_required=False the invite stores 0 but the hash check still ran, making the invite unredeemable with any pin other than the random one mint() generated. Now gate the PIN verification on row['pin_required'] being truthy. N1 fix: 3 of 5 B1 collab regression tests only tested mint storage and never exercised the redeem path — they could not detect the A2 bug. Updated test_mint_collab_invite, test_mint_collab_invite_no_pin_required, and test_default_kind_is_agent to also verify redeem correctness. Added test_mint_redeem_pin_not_required_e2e: mints with pin_required=False and successfully redeems with empty string — kills A1+A2 together. Store tests: 41/41 pass.
5503bac to
877282a
Compare
jaylfc#2048) Fix three regressions introduced by 877282a: - Restore invite["record"]["invite_id"] (was broken to invite["id"]) - Restore pin_required=False as top-level mint arg (was moved to metadata) - Restore cascade fail-closed guard (was weakened to skip on None store) Add set_sponsor guard: never re-parent an already-sponsored identity (the guard is now in production code, not reimplemented in tests). Fix get_project_setting: handle non-dict settings (malformed DB row). Rewrite N1 regression tests through real production paths: - TestMintReturnShape: exercises process_delegation_request e2e - TestSetSponsorGuard: tests production set_sponsor guard directly - Add TestDelegationE2E: mint + redeem e2e (kills A1+A2 together) 31/31 tests pass (D1 + invite_store).
|
Fixes pushed to 035e722. Summary of corrections: A2 (PIN skip):
Guard added to
N1 tests rewritten through real production paths:
Tests: 22/22 (D1 delegation) + 41/41 (invite_store) + 12/12 (project_store) + 124/124 (agent_registry) = 199 passed. |
|
@coderabbitai review |
✅ Action performedReview finished.
|
|
Verified all eight fix claims in your 16:06 summary against 035e722 at code level, including running the rewritten tests against the regressed commit (6/6 correctly fail there, pass at head). All eight are true, and the summary's scoping was honest about what it did not claim. This is the reporting standard; the 2182 comment earlier today was the opposite. State: the minimum-to-mergeable bar defined in my earlier review (A2 pin-skip + mint-to-redeem E2E) is met. Merge happens when a real bot review lands on this head (CodeRabbit re-triggered; its limit had eaten the last one). The A3/B1/B2/C items from the after-that list stay open as immediate follow-ups, with one flag raised in priority: A3 means the only reachable production path mints an invite whose ID is never delivered, so D1 cannot complete cross-instance end to end until A3 lands. Card follows on the board; treat it as next after the current batch. |
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
tinyagentos/projects/invite_store.py (1)
473-481: 🩺 Stability & Availability | 🟡 Minor | ⚡ Quick winFailed JSON parse leaves
metadata/scopesas a raw string.Callers treat these as
dict/list(e.g. the invite-redemption path readsmetadata["sponsor_contact_id"]), so a malformed row surfaces as anAttributeErrorrather than a benign default.🛡️ Proposed fix
- for field in ("scopes", "metadata"): - if field in d and isinstance(d[field], str): - try: - d[field] = json.loads(d[field]) - except (ValueError, TypeError): - pass + for field, fallback in (("scopes", []), ("metadata", {})): + if field in d and isinstance(d[field], str): + try: + d[field] = json.loads(d[field]) + except (ValueError, TypeError): + d[field] = fallback🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tinyagentos/projects/invite_store.py` around lines 473 - 481, Update the row deserialization logic around the metadata and scopes fields so failed JSON parsing replaces the raw string with the appropriate benign default for each field: a dictionary for metadata and a list for scopes. Preserve successfully decoded JSON values and return the normalized row from the existing deserialization function.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@tinyagentos/projects/invite_store.py`:
- Around line 473-481: Update the row deserialization logic around the metadata
and scopes fields so failed JSON parsing replaces the raw string with the
appropriate benign default for each field: a dictionary for metadata and a list
for scopes. Preserve successfully decoded JSON values and return the normalized
row from the existing deserialization function.
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 8f3d19e2-3e48-44aa-be31-343c0d2d19c2
📒 Files selected for processing (6)
tests/projects/test_invite_store.pytests/test_collab_d1_delegation.pytinyagentos/agent_registry_store.pytinyagentos/delegation_handler.pytinyagentos/projects/invite_store.pytinyagentos/projects/project_store.py
🚧 Files skipped from review as they are similar to previous changes (3)
- tinyagentos/projects/project_store.py
- tinyagentos/agent_registry_store.py
- tinyagentos/delegation_handler.py
|
Discharged the delivery-path merge-blocker (head
Regression tests now exercise the delivery path directly ( |
| # and the receiver 403s it. | ||
| envelope = build_envelope( | ||
| from_username=from_username, | ||
| to_username=to_contact_id.removeprefix("hub:"), |
There was a problem hiding this comment.
SUGGESTION: removeprefix("hub:") silently no-ops on non-hub: contact ids
The new code assumes every to_contact_id is a hub:<user> id. If a legacy or alternate contact id (e.g. peer:abc) is passed, removeprefix is a no-op and the bare id is forwarded to build_envelope, which will then prefix it again and produce an envelope addressed to hub:peer:abc — never matching the remote local_id and silently 403'd by the recipient. Either normalise at the caller (strip the hub: once and document the contract) or raise a clear ValueError here so the caller fixes the wrong shape rather than discovering it via a missing delivery.
| outbound_token = peer_link.get("outbound_token", "") | ||
| # Endpoints are plain URL strings (the canonical shape written by | ||
| # ``establish_peer_link``), not ``{"url": ..., "priority": ...}`` dicts. | ||
| for ep in peer_link["endpoints"]: |
There was a problem hiding this comment.
SUGGESTION: Endpoint priority sort was silently dropped
The previous code sorted endpoints by e.get("priority", 99) so the lowest-priority-number endpoint (the one the contact declared as primary) was always tried first. The replacement iterates peer_link["endpoints"] in dict-iteration order, which under SQLite aiosqlite is row order — non-deterministic across queries and changing the wire shape means a healthy primary endpoint can sit behind a dead one on every retry. Either keep a deterministic order (sort by presence of priority, fall back to insertion order) or document the new ordering semantics explicitly.
| } | ||
|
|
||
| try: | ||
| decision = await decision_store.create( |
There was a problem hiding this comment.
WARNING: _handle_delegation_status is not idempotent on invite_id
The handler unconditionally calls decision_store.create(...) for every received envelope. Nonces protect against exact replays, but nothing prevents two distinct delegation_status envelopes (different nonces) for the same invite_id — e.g. a retry from the sponsor after a 5xx they treated as failed, or the sender looping the delivery path. The result is two free_text decisions with identical metadata.invite_id and identical questions, polluting the user's Decisions feed and potentially confusing the agent runner polling for "the" delegation decision. Check decision_store.list(metadata_invite_id=invite_id) (or extend the store with a metadata JSON-key lookup) before creating and return the existing decision_id on duplicates.
| } | ||
|
|
||
| try: | ||
| decision = await decision_store.create( |
There was a problem hiding this comment.
SUGGESTION: Decision row is created without project_id or user_id
body_data["project_id"] is read into metadata but never passed as the project_id= kwarg to decision_store.create, so the row's project_id column is NULL. Any project-scoped query on the recipient side (e.g. an agent runner listing project_id == X decisions) will not surface this delegation notification. Same for user_id — the row is ownerless, so only an admin (or no one, depending on the answer route's auth) can act on it. For consistency with _handle_collab_invite at line 553, either pass project_id=project_id or None to create, or at minimum document why the decision is intentionally project-less.
|
|
||
| try: | ||
| decision = await decision_store.create( | ||
| from_agent=f"peer:{contact_id}", |
There was a problem hiding this comment.
SUGGESTION: from_agent=f"peer:{contact_id}" mis-attributes the notification
The decision is a system-generated notification about a delegation approval, not a request from the peer contact. Tagging it with from_agent="peer:<contact>" makes it look like the remote contact is asking the local user a free_text question. Downstream consumers that filter by from_agent prefix (e.g. surfacing "questions from peers" in the UI) will mis-categorise this. Use a system-prefixed identity (e.g. from_agent="system:delegation") or the local user's identity, and keep the originating contact only in metadata.contact_id.
| async with self._db.execute( | ||
| "SELECT * FROM decisions " | ||
| "WHERE json_extract(metadata, ?) = ? ORDER BY created_at DESC", | ||
| (f"$.{key}", value), |
There was a problem hiding this comment.
SUGGESTION: key is interpolated into the JSON path via f-string — f"$.{key}". The caller controls this string and SQLite JSON1 will reject invalid path syntax with an error rather than allowing injection, but the API surface still trusts the caller to pass a safe metadata key. Validate key (e.g. key.isidentifier() or re.match(r"^[A-Za-z_][A-Za-z0-9_]*$", key)) at function entry so future callers cannot break the query or leak schema details via a malformed JSON path.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| """ | ||
| async with self._db.execute( | ||
| "SELECT * FROM decisions " | ||
| "WHERE json_extract(metadata, ?) = ? ORDER BY created_at DESC", |
There was a problem hiding this comment.
SUGGESTION: find_by_metadata is documented as the idempotency probe for _handle_delegation_status, but the query has no LIMIT. A pathological store with thousands of rows sharing one invite_id (which the schema does not prevent) would force the whole list through _row_to_decision and json.loads on every duplicate delivery. Add LIMIT 1 and drop ORDER BY created_at DESC — the caller already takes existing[0], and the metadata index is keyed by invite_id so order does not matter for the duplicate-detection semantic.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| project = await project_store.get_project(project_id) | ||
| if project: | ||
| user_id = project.get("user_id") or "" | ||
| except Exception: |
There was a problem hiding this comment.
SUGGESTION: The new try/except Exception around project_store.get_project(project_id) silently swallows every error to user_id = "". Inconsistent with the sibling block just above (L472) which logs a warning on find_by_metadata failure, and inconsistent with the new create block below (L522) which logs an error. A ProjectStore bug that makes get_project raise will hide the owner of every delegation-status decision in this inbox. Add logger.warning(...) here for parity.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
| invite_id, exc, | ||
| ) | ||
| existing = [] | ||
| if existing: |
There was a problem hiding this comment.
SUGGESTION: Idempotency is keyed on invite_id alone, not (invite_id, project_id). If two sponsors independently chose the same invite_id (or a manual DB edit did), the second delivery would be silently swallowed and returned as a duplicate of the first project's decision — leaking the other project's decision_id in the response and starving the actual recipient. Either tighten the lookup to json_extract(metadata, '$.invite_id') = ? AND json_extract(metadata, '$.project_id') = ?, or document that invite_id is project-scoped.
Reply with @kilocode-bot fix it to have Kilo Code address this issue.
…ork, jaylfc#2048) Replace the single sponsor_contact_id column on agent_registry with an agent_sponsorship association keyed by (canonical_id, project_id), so one agent identity reused by handle across projects can be sponsored independently by each project's delegating contact. - v8 migration backfills the association from the legacy sponsor_contact_id column (keyed to the empty project_id sentinel), reproducing the old single-sponsor cascade for pre-change rows. - list_by_sponsor resolves through the association. - set_sponsorship writes INSERT OR IGNORE (atomic first-writer-wins on the PRIMARY KEY); remove_sponsorship + list_sponsorships_for_identity / list_sponsorships_by_contact added. - cascade_sponsor_revoke removes the contact's association rows and revokes the identity only when no association remains. - approve_request_record records sponsorship via the association instead of a column stamp.
…project) association, jaylfc#2048) Cover jaylfc's acceptance criteria: - two contacts / two projects / one handle: A's revoke spares the identity under B's sponsorship; B's revoke kills it outright. - project-scoped revoke spares other projects. - per-contact kill-switch observable through agent_token_auth._verify_agent_scope refusing a live bearer (403), not list_by_sponsor row count. - concurrency: two writes for the same (identity, project) leave exactly one row with a deterministic first-writer-wins winner. - v8 migration backfill reproduces the old single-sponsor cascade on a pre-change row. Replaces the old TestSetSponsorAtomicGuard 'in (hub:A, hub:B)' assertion.
|
Sponsorship rework landed (head What changed:
Acceptance (all four named cases):
Tests: |
…des (D1) Implement cross-user agent delegation (milestone D1 of jaylfc#2012): - Add sponsor_contact_id column to agent_registry (schema + migration + store API) - Delegation-request envelope dispatch in peer inbox (kind=delegation_request) - Policy gate with per-project auto_approve_delegation knob (default OFF) - Sponsored scope tiers: {a2a_send, a2a_receive, project_tasks, canvas_read, registry_feeds_read} - Hard-deny files_write and decisions_write at scope validation - Decisions card integration: collab_delegation_gate kind triggers invite mint on approve - Revoke cascade: contact/membership change → revoke sponsored identities → unassign tasks - 3-level kill-switch: per-agent (existing), per-contact pause, per-instance panic - Invite metadata column with sponsor_contact_id propagation through redeem flow - Project store helpers: is_project_member(), get/set_project_setting() Fixes: jaylfc#2019
CRITICAL: Restore reply assignment in _apply_delegation_grant (regression from collapsed if/elif/else) WARNING: cascade_sponsor_revoke now filters agents by project_id via project_store.is_project_member for membership-scoped revoke WARNING: complete_delegation_approval re-applies scope denylist (validate_delegation_scopes) before minting invite SUGGESTION: Add kill_switch_reenable() for per-instance panic recovery SUGGESTION: A2A audit line emitted for contact-wide revoke across all projects the contact is a member of
WARNING: Re-validate project membership at approval time in complete_delegation_approval (fail-closed runtime check per design spec 5) WARNING: Remove fragile list_projects_for_member call (method does not exist on project_store). Simplify A2A audit to project-scoped only. SUGGESTION: Route error answer back when complete_delegation_approval raises, so the remote contact receives feedback on approval failure.
A2 fix (jaylfc#2048): redeem() always validated the PIN hash regardless of pin_required. When pin_required=False the invite stores 0 but the hash check still ran, making the invite unredeemable with any pin other than the random one mint() generated. Now gate the PIN verification on row['pin_required'] being truthy. N1 fix: 3 of 5 B1 collab regression tests only tested mint storage and never exercised the redeem path — they could not detect the A2 bug. Updated test_mint_collab_invite, test_mint_collab_invite_no_pin_required, and test_default_kind_is_agent to also verify redeem correctness. Added test_mint_redeem_pin_not_required_e2e: mints with pin_required=False and successfully redeems with empty string — kills A1+A2 together. Store tests: 41/41 pass.
jaylfc#2048) Fix three regressions introduced by 877282a: - Restore invite["record"]["invite_id"] (was broken to invite["id"]) - Restore pin_required=False as top-level mint arg (was moved to metadata) - Restore cascade fail-closed guard (was weakened to skip on None store) Add set_sponsor guard: never re-parent an already-sponsored identity (the guard is now in production code, not reimplemented in tests). Fix get_project_setting: handle non-dict settings (malformed DB row). Rewrite N1 regression tests through real production paths: - TestMintReturnShape: exercises process_delegation_request e2e - TestSetSponsorGuard: tests production set_sponsor guard directly - Add TestDelegationE2E: mint + redeem e2e (kills A1+A2 together) 31/31 tests pass (D1 + invite_store).
…ites key - Rebased onto current origin/dev (no conflicts) - Fixed app.state key mismatch: project_invite_store → project_invites (app.py registers as project_invites, handler was looking up the wrong key) - invite["record"]["invite_id"] already correct in both call sites
…project_invite_store)
CodeRabbit findings (PR jaylfc#2048): - CRITICAL: Both mint call sites correctly use invite['record']['invite_id']. - MAJOR: Both _auto_approve_delegation and complete_delegation_approval passed invalid metadata={...} dict to mint(). Fix: drop metadata, promote display_name to top-level kwarg.
…uto-approve setting (jaylfc#2048) The manual approval path had no caller: complete_delegation_approval (the 'minted on decisions approve' handler) was never invoked, and the auto_approve_delegation setting was unexposed (set_project_setting had no route). Wire _apply_collab_delegation_grant into answer_decision so approving a collab_delegation_gate decision actually mints the sponsored invite, and add GET/PUT project setting routes for auto_approve_delegation.
…ment (jaylfc#2048) The rebase onto dev accidentally reverted five CI workflow files: it deleted secret-ignores-gate.yml and store-wiring-gate.yml, downgraded node 22 -> 20 in ci.yml, dropped the deleted-symbols-gate "edited" retrigger, and reverted security.yml's uv/pip-audit hardening. Restore all five to dev's version — the delegation feature does not touch CI. Docs-Reviewed: delegation is a new cross-user agent surface. The D1 contract is specified in the PR body and pinned by 15 tests in tests/test_collab_d1_delegation.py; agent-coordination.md and agent-manual do not yet cover delegation, and a full doc pass should follow this slice rather than block it.
…, fail-closed re-check, TOCTOU set_sponsor, task release, high-entropy invite IDs (jaylfc#2048)
…d + tighten scope allowlist BLOCKING: Approved delegation invites never reached the requester. - Add send_envelope() to peer.py for reusable outbound peer delivery. - Add delegation_status envelope kind on both outbound (peer.py) and inbound (routes/peer.py) sides. - Wire invite_id through _apply_collab_delegation_grant: after mint succeeds, deliver delegation_status envelope to requester via the peer channel so their agent runner can redeem the invite. MEDIUM: set_sponsor guard prevented cross-project re-parenting. - Allow re-parenting to a DIFFERENT sponsor (handle-based reuse across projects). The old guard blocked any change, so an identity first sponsored by contact A that joined B's project stayed stamped A, making cascade_sponsor_revoke(B) miss it. - Set-same-sponsor is still a no-op; clearing (None) still works. - Caller in agent_auth_requests.py now handles None return. LOW: Unknown delegation scopes rode through to the mint. - Replace permissive elevated-scope logging with VALID_SCOPES gate in _validate_delegation_envelope_body — unknown scopes now 400 instead of silently passing through bundled approve_deny cards. Tests: test_set_sponsor_reparents_to_different_sponsor (replaces old immutability test), test_set_sponsor_noop_on_same_sponsor (new).
…wance The old test asserted that re-parenting was blocked; now it asserts that re-parenting succeeds (cross-project identity reuse via handle).
…b:hub:<user> build_envelope prefixes 'hub:' itself; send_envelope was passing the already prefixed contact id through as to_username, producing a hub:hub:<user> envelope that the receiver 403s. Strip the prefix before building the envelope.
establish_peer_link stores plain URL strings (endpoints: list[str]), but
send_envelope iterated them as {'url', 'priority'} dicts, so sorted(...)
raised AttributeError outside its own try. Treat endpoints as strings.
…edeem The receiver of a delegation_status envelope only logged and echoed the invite_id; nothing persisted it, so the requester's agent runner had no record to poll. Persist a decision with invite_id in metadata (mirroring collab_invite) and drop the now-unused to_username in _deliver_delegation_invite.
send_envelope / _deliver_delegation_invite had zero coverage, which is why the double-prefix and endpoint-shape bugs survived green. Add regression tests that drive the real outbound path against a stub inbox (delivered is True, envelope addressed hub:<user>) and assert delegation_status persists a durable decision.
…ic endpoint ordering send_envelope now raises ValueError when to_contact_id is not a hub:<user> id (previously removeprefix no-op'd and produced a hub:peer:abc envelope that the receiver silently 403s). Endpoint iteration is deterministic: plain URL strings keep their declaration order, and dict endpoints (with an optional priority) are sorted lowest-priority-first so the declared primary endpoint is always tried before a fallback.
…perly - Add DecisionStore.find_by_metadata(key, value) (json_extract lookup) and use it so a retried delegation_status envelope for the same invite_id returns the existing decision instead of minting a duplicate. - Pass project_id and user_id (project owner) through to create so the row is project-scoped and owned, not NULL/ownerless. - Tag the notification from_agent="system:delegation" (a system notification about an approval, not a request from the peer); originating contact stays in metadata.contact_id.
- idempotency: retried delegation_status returns the existing decision_id without a duplicate row. - project scoping + attribution: decision carries project_id, owner user_id, and from_agent=system:delegation (contact kept in metadata). - send_envelope: non-hub contact ids raise ValueError; dict endpoints are tried lowest-priority-first. - DecisionStore.find_by_metadata: exact JSON-key match, no substring false positives.
…ork, jaylfc#2048) Replace the single sponsor_contact_id column on agent_registry with an agent_sponsorship association keyed by (canonical_id, project_id), so one agent identity reused by handle across projects can be sponsored independently by each project's delegating contact. - v8 migration backfills the association from the legacy sponsor_contact_id column (keyed to the empty project_id sentinel), reproducing the old single-sponsor cascade for pre-change rows. - list_by_sponsor resolves through the association. - set_sponsorship writes INSERT OR IGNORE (atomic first-writer-wins on the PRIMARY KEY); remove_sponsorship + list_sponsorships_for_identity / list_sponsorships_by_contact added. - cascade_sponsor_revoke removes the contact's association rows and revokes the identity only when no association remains. - approve_request_record records sponsorship via the association instead of a column stamp.
…project) association, jaylfc#2048) Cover jaylfc's acceptance criteria: - two contacts / two projects / one handle: A's revoke spares the identity under B's sponsorship; B's revoke kills it outright. - project-scoped revoke spares other projects. - per-contact kill-switch observable through agent_token_auth._verify_agent_scope refusing a live bearer (403), not list_by_sponsor row count. - concurrency: two writes for the same (identity, project) leave exactly one row with a deterministic first-writer-wins winner. - v8 migration backfill reproduces the old single-sponsor cascade on a pre-change row. Replaces the old TestSetSponsorAtomicGuard 'in (hub:A, hub:B)' assertion.
a61324c to
eb80bf9
Compare
|
@- |
2 participants
Task: t_b72392f4. Fixes #2019.
Summary
Implement cross-user agent delegation (milestone D1 of EPIC #2012).
Changes
Files changed
Tests
311 tests pass across: test_collab_d1_delegation, test_agent_registry, test_project_store, test_invite_store, test_contacts_peer, test_auth_middleware, test_auth_requests_store
Summary by CodeRabbit