Activity

1. senamakel

   assigned

   oxoxDev

   on Nov 9, 2026
2. tinysweeper

   added

   priority: p3Whenever. Cosmetic, a nicety, or a cleanup with no user visible effect.Whenever. Cosmetic, a nicety, or a cleanup with no user visible effect.

   on Nov 9, 2026
3. tinysweeper commented on Nov 9, 2026

   

   tinysweeperBot

   on Nov 9, 2026

   More actions

   Tracking issue for a one-off triage pass over the repo's open PRs and issues (5 PRs, 6 issues) against main, with per-item triage comments and a summary. This is process/coordination work, not a bug or feature; low urgency since it is a scheduled housekeeping milestone task.

   Labelled priority: p3.
4. tinysweeper

   added the

   Task

   issue type on Nov 9, 2026
5. M3gA-Mind

   assigned

   M3gA-Mind

   and unassigned

   oxoxDev

   on Nov 11, 2026
6. M3gA-Mind commented on Nov 11, 2026

   

   M3gA-Mind

   on Nov 11, 2026

   Collaborator

   More actions

   Status update, 2026-11-11 14:50 IST: merged #148. The tables below are the triage as it was done; rows for those items are now settled.

   tinybox: open PR triage (tinybox#151, team#45)

   Official main 7146de6. openhuman pins 8febf22 = v0.2.5: 6 commits behind main, no source or contract change between them (sha2 bump only), contract (2, 3) on both. Tested on our CI runner: #150, #149 and #148, each merged onto main. Main's Windows feature-tests fails on the run_native_shell_* tests (run 42013264157), and the three dependabot PRs fail only that job.

   PR	Verdict	Reason	Next step
   #150 base64 0.22.1→0.23.1	ready to merge	Clean merge; --locked build and workspace tests pass (1515/0). The only red is main's Windows flake.	Merge (re-run or admin over the flake).
   #149 tokio 1.53.3→1.54.1	ready to merge	Lock-only; build and tests pass (1515/0). The only red is the same Windows flake family.	Merge.
   #141 actions/checkout 4.4.0→7.0.1	ready to merge	The new SHA is exactly tag v7.0.1, still pinned; the jobs using it pass. Workflow only.	Merge.
   #148 retained agent37 boxes (contract 2.4)	MERGED 11-11 (690bb56), P2 still open	CF green (build, clippy, 631 tests). One codex P2 is unanswered: closing one of two resources sharing an adopted box deletes the shared record. agent37 is live in openhuman, and its 4 CreateRequest literals break on the re-pin.	File a follow-up issue for the codex P2 (closing one shared retained box strands the other); plan the openhuman re-pin, module record and literals together.
   #84 custom Rust homes (fork)	needs changes	Still a real gap on main, but 785 behind, 2 conflicts, written for contract 1.20 (main is 2.3), last run red on Linux and macOS; author quiet since 10-22.	Author rebases or a fresh smaller PR; maintainer decides whether to close.

   Issues

   By code reading; identical at main 7146de6 and the pin 8febf22 (jail/module/host crates unchanged between them).

   Issue	Verdict	Reason	Next step
   #109 bwrap allow_net: isolated networking (pasta/slirp4netns)	still real (p2)	bwrap/mod.rs:352-367 only reports Kernel when !allow_net; no pasta/slirp4netns; Linux registry MCP servers were refused (EX63 run 39292268284, per the issue)	tinybox PR per the proposal; first confirm openhuman still sets allow_net + require_process_isolation
   #104 bwrap follow-ups from #99 review	still real (p3)	none of the 6 items done: bwrap/mod.rs:64 still says fd binds need 0.10+; launcher.rs base_dir isn't canonicalized; no seccomp	one security PR first (fd-bound grants via openat2, canonical base_dir, tests); track seccomp/limits separately; related to #84's grants
   #89 Seatbelt deny_read: scratch carve-out + name/pattern denials	still real (p2)	macos.rs:236-242 renders the deny block last with no scratch carve-out; native_shell/mod.rs:109-110 adds the scratch after prepare; denials are exact paths only	carve-out first (scratch before prepare, or re-allow it last), patterns as a separate contract change; the 0.1.36 target is stale; same grant path as #84
   #27 sandbox-only tracker	still real (corrected)	the contract types for every phase are on main, but no backend enforces NetworkPolicy::Allowlist, nothing emits SandboxEvent::Denied, and bwrap has no limits; the host sends resources: None from all 3 builders; openhuman#7328 is closed	narrow it to the tinybox enforcement gaps; host work → new openhuman issue (draft: tinybox-issue-27-host-draft.md)
   #25 bound the Seatbelt test's gc.pid wait	still real (p3)	macos_tests.rs:382-389 still loops with no deadline	small test PR: 5s deadline, kill and reap, clear failure message
   #22 Windows AppContainer backend can't be enabled	already fixed (#140, via fork #66)	windows.rs is now compiled on Windows (lib.rs:70-71); the AppContainer backend spawns via a crate-owned child (spawn_owned_captured); CI fences the unsafe	confirm the Windows default backend uses the owned path, then close

   Update 2026-11-11 14:30 IST:

   • tinybox main moved to b498faf (Retained agent37 boxes: sleep when idle, adopt by owner key, expire; wait for waking instances #148 merged as 690bb56, then the 0.2.6 release chore(release): tinybox 0.2.6 #152).
   • module-ci / supply-chain is now red on main and on re-runs: RUSTSEC-2026-0459 (libsqlite3-sys 0.38.2, C SQLite CVE-2026-119375). It comes via tinystoragedrivers 2.1.2 (rev 6226b9484c) → rusqlite 0.40.2, so the fix is a tsd release on rusqlite 0.41, then a tinybox tsd pin bump.
   • Both supply-chain and Windows feature-tests are REQUIRED checks (main's ruleset). Until those two are fixed, deps: bump base64 from 0.22.1 to 0.23.1 #150, deps: Bump tokio from 1.53.3 to 1.54.1 in the cargo-minor-and-patch group across 1 directory #149 and ci: Bump actions/checkout from 4.4.0 to 7.0.1 in the github-actions group across 1 directory #141 can only merge as admin merges, even though they're otherwise ready.