lhumina_code/hero_skills
4
0
Fork 0
Code Issues 46 Pull requests Projects Releases 2 Packages 1 Wiki Activity Actions

Per-repo cleanup blocking lab-publish CI on 14 of 31 wired repos (s127 follow-up to #268) #269

New issue
Open
opened 2026-05-19 22:09:37 +00:00 by mik-tf · 1 comment
mik-tf commented 2026-05-19 22:09:37 +00:00
Owner
Copy link

Summary

s127 wired the canonical lab-publish.yaml workflow into 31 of 35 D-07 demo-set repos via hero_skills#268. 14 of the 31 are now auto-publishing fresh musl-x86_64 binaries to releases/tag/latest on every push to development. The other 17 trip on pre-existing latent bugs in their own source / Cargo.toml — the workflow file itself is correct (proven on the 14 green repos across 1-binary / 3-binary / multi-bin workspace shapes). When each repo's per-repo bug is fixed, the next push to development auto-goes-green; no central tracker or re-wiring needed.

This issue catalogs the 17 failures by category + fix surface, so repo owners can pick up their own cleanup independently.

Status after s127

Workflow file landed in: 31 repos (every D-07 35-set member except hero_voice / hero_os / hero_office / mycelium_network — all deferred for separate session-level decisions).

Currently green (binaries auto-publishing): 14 — hero_skills, hero_proc, hero_router, hero_db, hero_proxy, hero_osis, hero_livekit, hero_biz, hero_whiteboard, hero_indexer, hero_code_indexer, hero_researcher, hero_webbuilder, hero_website_framework.

Currently red (pre-existing bug): 17 — see categories below.

Special-case deferred (no workflow yet): 4 — hero_voice (ONNX, needs gnu per D-05), hero_os (WASM, separate target), hero_office (private repo, auth quirk), mycelium_network (development_crate_layout branch, geomind_code).

Failure categories

1. Path-dep hero_admin_lib to hero_website_framework (6 repos)

Fix surface: per-repo Cargo.toml flip path = "../../../hero_website_framework/crates/hero_admin_lib" → git = "https://forge.ourworld.tf/lhumina_code/hero_website_framework". Pattern already landed in mycelium_admin in s125.

  • hero_code
  • hero_codescalers
  • hero_collab
  • hero_lib_rhai (also hero_foundry_core path-dep)
  • hero_matrixchat
  • hero_planner

2. Lab post-process < 512KB binary-size guard (4 repos → FIXED in same session via env-var override)

Fix landed in s127: HERO_MIN_BIN_BYTES=0 env var added to canonical workflow; lab's check is now overridable.

  • hero_aibroker (mcp_ping = 444KB)
  • hero_compute (hero_compute_server = 315KB, my_compute_mos_admin = 311KB)
  • hero_lib (herolib_web = 522KB — 510 KiB, just under)
  • hero_rpc (hero_rpc_generator = 428KB; ALSO has issue #5 below)

These should re-go-green on next push after the s127 lab + workflow update lands.

3. C++ cross-compile (x86_64-linux-musl-g++ missing) (2 repos)

Fix surface: Debian's musl-tools package only ships musl-gcc (C only), not musl-g++. Three viable approaches:

  • Per D-05: matrix-swap these repos to gnu target via --platforms linux-x86_64 (workflow variant)

  • OR builder container with musl-cross toolchain pre-installed (PR to despiegk's image)

  • OR clang + --target=x86_64-unknown-linux-musl (workflow change, may need libc++ headers)

  • hero_embedder (esaxx-rs C++ — already covered by D-05 lock)

  • hero_memory (C++ vector-backend dep)

4. Cargo feature spec error (2 repos)

Fix surface: per-repo Cargo.toml.

  • hero_agent — reqwest feature rustls not selectable on resolved 0.12 (the dep uses dep:rustls syntax which doesn't auto-create the implicit feature). Switch to rustls-tls or expose the feature explicitly.
  • hero_wallet — askama feature with-axum not available on 0.16 (renamed/removed in 0.13+).

5. Cargo links = collision (1 repo)

Fix surface: per-repo Cargo.toml dedup.

  • hero_foundry — two versions of rusqlite in the dep tree (0.31 via r2d2_sqlite^0.24 and 0.39 directly) both link to sqlite3 native. Pin one or remove the older r2d2_sqlite chain.

6. musl portability (__rlimit_resource_t glibc-only) (1 repo)

Fix surface: per-repo source cleanup. s70 had the same fix landed but it regressed.

  • hero_logic — some setrlimit-style code references __rlimit_resource_t. Replace with the libc::Resource enum or cfg-gate.

7. Unresolved git merge-conflict marker in source (1 repo)

Fix surface: per-repo source cleanup. Single grep + commit.

  • hero_slides — hero_slides_lib has <<<<<<< / >>>>>>> markers in source (cargo error: "encountered diff marker").

8. Stale herolib_core API references (1 repo)

Fix surface: per-repo cargo update -p herolib_core + rebuild.

  • hero_books — source calls herolib_core::base::hero_socket_dir and path_var, but the pinned herolib_core version doesn't have those yet (or has them removed).

9. Stale service.toml package references (1 repo)

Fix surface: per-repo service.toml cleanup.

  • hero_rpc — hero_recipes_server / hero_recipes_admin listed in service.toml but the crates don't exist (orphan refs from rename/removal).

Cleanup pattern

For each failing repo:

cd lhumina_code/<repo>
git fetch origin && git checkout development && git pull --ff-only
git checkout -b development_mik
# fix the per-category issue
git add -A && git commit -m "fix: <category>"
git checkout development && git merge --squash development_mik
git commit -m "fix: <category>" && git push origin development
git branch -D development_mik

The next push automatically triggers the already-wired lab-publish.yaml, which goes green if the per-repo bug is the only blocker.

Out of scope

  • hero_voice / hero_os / hero_office / mycelium_network — 4 special-case repos that need workflow variants (gnu for ONNX, WASM build, private auth, branch override). Track separately.
  • The lab post-process < 512KB guard semantics. Current fix is an env-var opt-out. Long-term the check should probably move from static size threshold to a strip/file-format sanity check, OR per-binary opt-out via service.toml. Tracked as a lab refactor candidate; not blocking any current work.

See also

  • hero_skills#268 — s127 origin issue (this issue is its post-rollout follow-up)
  • s125 mycelium_network — the precedent path-dep → git-dep flip pattern in mycelium_admin/Cargo.toml
  • D-05 — locks ONNX services on gnu, not musl (governs hero_voice / hero_embedder / hero_memory matrix decision)
## Summary s127 wired the canonical `lab-publish.yaml` workflow into 31 of 35 D-07 demo-set repos via [hero_skills#268](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/268). **14 of the 31 are now auto-publishing fresh musl-x86_64 binaries to `releases/tag/latest` on every push to `development`.** The other 17 trip on pre-existing latent bugs in their own source / Cargo.toml — the workflow file itself is correct (proven on the 14 green repos across 1-binary / 3-binary / multi-bin workspace shapes). When each repo's per-repo bug is fixed, the next push to `development` auto-goes-green; no central tracker or re-wiring needed. This issue catalogs the 17 failures by category + fix surface, so repo owners can pick up their own cleanup independently. ## Status after s127 **Workflow file landed in:** 31 repos (every D-07 35-set member except hero_voice / hero_os / hero_office / mycelium_network — all deferred for separate session-level decisions). **Currently green (binaries auto-publishing):** 14 — hero_skills, hero_proc, hero_router, hero_db, hero_proxy, hero_osis, hero_livekit, hero_biz, hero_whiteboard, hero_indexer, hero_code_indexer, hero_researcher, hero_webbuilder, hero_website_framework. **Currently red (pre-existing bug):** 17 — see categories below. **Special-case deferred (no workflow yet):** 4 — hero_voice (ONNX, needs gnu per D-05), hero_os (WASM, separate target), hero_office (private repo, auth quirk), mycelium_network (`development_crate_layout` branch, geomind_code). ## Failure categories ### 1. Path-dep `hero_admin_lib` to `hero_website_framework` (6 repos) **Fix surface**: per-repo Cargo.toml flip `path = "../../../hero_website_framework/crates/hero_admin_lib"` → `git = "https://forge.ourworld.tf/lhumina_code/hero_website_framework"`. Pattern already landed in mycelium_admin in s125. - hero_code - hero_codescalers - hero_collab - hero_lib_rhai (also `hero_foundry_core` path-dep) - hero_matrixchat - hero_planner ### 2. Lab post-process < 512KB binary-size guard (4 repos → FIXED in same session via env-var override) **Fix landed in s127**: `HERO_MIN_BIN_BYTES=0` env var added to canonical workflow; lab's check is now overridable. - hero_aibroker (mcp_ping = 444KB) - hero_compute (hero_compute_server = 315KB, my_compute_mos_admin = 311KB) - hero_lib (herolib_web = 522KB — 510 KiB, just under) - hero_rpc (hero_rpc_generator = 428KB; ALSO has issue #5 below) These should re-go-green on next push after the s127 lab + workflow update lands. ### 3. C++ cross-compile (`x86_64-linux-musl-g++` missing) (2 repos) **Fix surface**: Debian's `musl-tools` package only ships `musl-gcc` (C only), not `musl-g++`. Three viable approaches: - Per D-05: matrix-swap these repos to gnu target via `--platforms linux-x86_64` (workflow variant) - OR builder container with `musl-cross` toolchain pre-installed (PR to despiegk's image) - OR clang + `--target=x86_64-unknown-linux-musl` (workflow change, may need libc++ headers) - hero_embedder (esaxx-rs C++ — already covered by D-05 lock) - hero_memory (C++ vector-backend dep) ### 4. Cargo feature spec error (2 repos) **Fix surface**: per-repo Cargo.toml. - hero_agent — `reqwest` feature `rustls` not selectable on resolved 0.12 (the dep uses `dep:rustls` syntax which doesn't auto-create the implicit feature). Switch to `rustls-tls` or expose the feature explicitly. - hero_wallet — `askama` feature `with-axum` not available on 0.16 (renamed/removed in 0.13+). ### 5. Cargo `links =` collision (1 repo) **Fix surface**: per-repo Cargo.toml dedup. - hero_foundry — two versions of `rusqlite` in the dep tree (0.31 via `r2d2_sqlite^0.24` and 0.39 directly) both link to `sqlite3` native. Pin one or remove the older `r2d2_sqlite` chain. ### 6. musl portability (`__rlimit_resource_t` glibc-only) (1 repo) **Fix surface**: per-repo source cleanup. s70 had the same fix landed but it regressed. - hero_logic — some `setrlimit`-style code references `__rlimit_resource_t`. Replace with the libc::Resource enum or cfg-gate. ### 7. Unresolved git merge-conflict marker in source (1 repo) **Fix surface**: per-repo source cleanup. Single grep + commit. - hero_slides — `hero_slides_lib` has `<<<<<<<` / `>>>>>>>` markers in source (cargo error: "encountered diff marker"). ### 8. Stale `herolib_core` API references (1 repo) **Fix surface**: per-repo `cargo update -p herolib_core` + rebuild. - hero_books — source calls `herolib_core::base::hero_socket_dir` and `path_var`, but the pinned herolib_core version doesn't have those yet (or has them removed). ### 9. Stale service.toml package references (1 repo) **Fix surface**: per-repo service.toml cleanup. - hero_rpc — `hero_recipes_server` / `hero_recipes_admin` listed in service.toml but the crates don't exist (orphan refs from rename/removal). ## Cleanup pattern For each failing repo: ```bash cd lhumina_code/<repo> git fetch origin && git checkout development && git pull --ff-only git checkout -b development_mik # fix the per-category issue git add -A && git commit -m "fix: <category>" git checkout development && git merge --squash development_mik git commit -m "fix: <category>" && git push origin development git branch -D development_mik ``` The next push automatically triggers the already-wired `lab-publish.yaml`, which goes green if the per-repo bug is the only blocker. ## Out of scope - hero_voice / hero_os / hero_office / mycelium_network — 4 special-case repos that need workflow variants (gnu for ONNX, WASM build, private auth, branch override). Track separately. - The lab post-process < 512KB guard semantics. Current fix is an env-var opt-out. Long-term the check should probably move from static size threshold to a strip/file-format sanity check, OR per-binary opt-out via service.toml. Tracked as a lab refactor candidate; not blocking any current work. ## See also - [hero_skills#268](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/268) — s127 origin issue (this issue is its post-rollout follow-up) - s125 mycelium_network — the precedent path-dep → git-dep flip pattern in mycelium_admin/Cargo.toml - D-05 — locks ONNX services on gnu, not musl (governs hero_voice / hero_embedder / hero_memory matrix decision)
mik-tf commented 2026-05-19 22:53:34 +00:00
Author
Owner
Copy link

s128 close-out — Surface A swept; 17→25 actually-shipping, two diagnoses redirected, two follow-up issues filed

Per-row outcomes

Category 1 — Path-dep hero_admin_lib (6 repos)

5 of 6 flipped to git-dep against hero_website_framework (s125 mycelium_admin precedent line, verbatim):

repo commit
hero_codescalers 3bc4d07
hero_lib_rhai febfa42
hero_matrixchat 89a983a
hero_planner 231f268
hero_collab e833369

hero_code was already on the git-dep form (line 85 of its Cargo.toml). Its CI failure is unrelated to this category — releases/tag/latest already ships hero_code-linux-musl-x86_64 + hero_code_admin + hero_code_server. Closing it out of Category 1 with no further action this session.

Category 4 — Cargo feature spec (2 repos)

  • hero_agent → 51c6551 — reqwest feature "rustls" → "rustls-tls" (the canonical name; "rustls" was never a valid feature on reqwest 0.12/0.13).
  • hero_wallet → 2f601c7 — dropped askama "with-axum" feature; integration is already covered by separate askama_axum dep in the same manifest.

Category 5 — links="sqlite3" collision (1 repo)

hero_foundry → a4a9e5d. Root cause: lab's D-08 policy auto-bumps rusqlite 0.31 → 0.39, but no published r2d2_sqlite release supports rusqlite ≥ 0.39 (latest r2d2_sqlite 0.31 caps at rusqlite ^0.37). The auto-bump pulls two libsqlite3-sys versions into the resolved graph, both linking to sqlite3 native.

Fix uses the documented [package.metadata.hero_builder.dep_overrides] escape hatch (dependency_upgrade.rs:59-69) to hold rusqlite at 0.31. Remove when r2d2_sqlite catches up.

Category 6 — __rlimit_resource_t musl portability (1 repo)

hero_logic → a55deca. libc::__rlimit_resource_t is glibc-only; on musl, libc::setrlimit's resource arg is c_int. Added a target_env-gated type alias RlimitResource so both glibc and musl link cleanly. Verified locally with cargo check --target x86_64-unknown-linux-musl.

Category 7 — Unresolved merge markers (1 repo) — DEFERRED, filed hero_slides#71

hero_slides development HEAD does NOT compile. Commit 6e39d677 by @despiegk — chore: auto-commit local changes before pull (2026-05-19 17:53 +0200) — is a merge that committed 138 unresolved conflict markers across 46 regions in 10 files, including the binary entry's 42 markers (main.rs) and 30 in generator.rs. The two parents conflict on incompatible structural intents:

  • 45efb36 by @Casper — server module-layout refactor (file renames into jobs/ and rpc/)
  • 93a3c99 by @despiegk — herolib_ai API migration (ImageRef relocation, error variant renames, agent invocation rewrites)

Resolving requires picking the winner per concern; the original authors have the context. Filed hero_slides#71 with details + suggested unwinding procedure.

The §3 estimate of ≤ 10 min (based on raw marker count) underestimated by an order of magnitude. Per feedback_branch_cleanup_only_own_authored.md, I left it alone.

Category 8 — Stale herolib_core API (1 repo) — DIAGNOSIS WAS WRONG

hero_books. Phase B confirmed (and local lab build --release --platforms linux-musl-x86_64 --workspace confirmed mid-build) that:

  1. herolib_core::base::hero_socket_dir and path_var() ARE exported by current herolib_core (base/mod.rs:8-9 + paths.rs:54,86).
  2. hero_books builds cleanly under lab locally — hero_books_admin (1.97MB), hero_books_web (4.34MB), hero_books_server and hero_do_hero_books produced before the test timeout.
  3. releases/tag/latest already has hero_books_admin, hero_books_web, hero_do_hero_books musl-x86_64 assets shipping from a prior workstation upload.

The lab-publish.yaml CI fail is not a code issue. Likely: transient infrastructure / asset-upload step / runner pre-condition. Recommend triggering a fresh CI run and harvesting the actual log when the next push lands on hero_books.

Category 9 — Stale service.toml refs (1 repo) — DIAGNOSIS WAS WRONG; filed hero_skills#270

hero_rpc. hero_recipes_server and hero_recipes_admin crates DO exist (example/recipe_server/crates/{hero_recipes_server,hero_recipes_admin}/). The actual failure surface is a lab orchestrator bug with nested workspaces: lab's executor.rs:54 runs cargo build -p hero_recipes_server from repo_root, but repo_root's Cargo.toml excludes example/recipe_server — so cargo says "package ID specification did not match." The bin is correctly discovered by discover_binaries_in_repo (which walks per-workspace cargo metadata), but the executor doesn't carry the bin's workspace root through to current_dir.

Detailed root-cause analysis and fix shapes at hero_skills#270. This blocks any D-07 repo that has nested-workspace examples under git_root — hero_rpc is the confirmed instance.

Net coverage

#269 baseline at s127 close:  14 / 31 wired   (45%)
s128 fixes (CI in flight):    +9              ← 5 path-dep + 2 feature + 1 sqlite + 1 rlimit
s128 already-shipping:        +2              ← hero_code + hero_books (older uploads)

Projected after CI runs:      25 / 31 wired   (81%)
Projected against 35-set:     25 / 35 D-07    (71%)

Unblocked remaining (against the 35-set):

  • 4 special-case deferred (Surface B) — hero_voice, hero_embedder, hero_memory (gnu matrix-swap per D-05), hero_office (private auth), mycelium_network (branch override), hero_os (WASM). Tracked at #268/#269 Out of scope sections.
  • hero_slides — needs #71 merge reconciliation.
  • hero_rpc — needs #270 lab patch.

Followups filed

  • hero_slides#71 — broken merge state on development; needs despiegk + Casper.
  • hero_skills#270 — lab executor.rs CWD bug on nested workspaces; blocks hero_rpc CI.
## s128 close-out — Surface A swept; 17→25 actually-shipping, two diagnoses redirected, two follow-up issues filed ### Per-row outcomes #### Category 1 — Path-dep `hero_admin_lib` (6 repos) 5 of 6 flipped to git-dep against `hero_website_framework` (s125 mycelium_admin precedent line, verbatim): | repo | commit | |---|---| | `hero_codescalers` | [3bc4d07](https://forge.ourworld.tf/lhumina_code/hero_codescalers/commit/3bc4d07) | | `hero_lib_rhai` | [febfa42](https://forge.ourworld.tf/lhumina_code/hero_lib_rhai/commit/febfa42) | | `hero_matrixchat` | [89a983a](https://forge.ourworld.tf/lhumina_code/hero_matrixchat/commit/89a983a) | | `hero_planner` | [231f268](https://forge.ourworld.tf/lhumina_code/hero_planner/commit/231f268) | | `hero_collab` | [e833369](https://forge.ourworld.tf/lhumina_code/hero_collab/commit/e833369) | `hero_code` was already on the git-dep form (line 85 of its Cargo.toml). Its CI failure is unrelated to this category — `releases/tag/latest` already ships `hero_code-linux-musl-x86_64` + `hero_code_admin` + `hero_code_server`. Closing it out of Category 1 with no further action this session. #### Category 4 — Cargo feature spec (2 repos) - `hero_agent` → [51c6551](https://forge.ourworld.tf/lhumina_code/hero_agent/commit/51c6551) — `reqwest` feature `"rustls"` → `"rustls-tls"` (the canonical name; `"rustls"` was never a valid feature on reqwest 0.12/0.13). - `hero_wallet` → [2f601c7](https://forge.ourworld.tf/lhumina_code/hero_wallet/commit/2f601c7) — dropped askama `"with-axum"` feature; integration is already covered by separate `askama_axum` dep in the same manifest. #### Category 5 — `links="sqlite3"` collision (1 repo) `hero_foundry` → [a4a9e5d](https://forge.ourworld.tf/lhumina_code/hero_foundry/commit/a4a9e5d). Root cause: **lab's D-08 policy auto-bumps rusqlite 0.31 → 0.39**, but no published `r2d2_sqlite` release supports rusqlite ≥ 0.39 (latest r2d2_sqlite 0.31 caps at rusqlite ^0.37). The auto-bump pulls two `libsqlite3-sys` versions into the resolved graph, both linking to `sqlite3` native. Fix uses the documented `[package.metadata.hero_builder.dep_overrides]` escape hatch ([dependency_upgrade.rs:59-69](https://forge.ourworld.tf/lhumina_code/hero_skills/src/branch/development/crates/lab/src/builder/dependency_upgrade.rs#L59-L69)) to hold rusqlite at 0.31. Remove when r2d2_sqlite catches up. #### Category 6 — `__rlimit_resource_t` musl portability (1 repo) `hero_logic` → [a55deca](https://forge.ourworld.tf/lhumina_code/hero_logic/commit/a55deca). `libc::__rlimit_resource_t` is glibc-only; on musl, `libc::setrlimit`'s resource arg is `c_int`. Added a `target_env`-gated type alias `RlimitResource` so both glibc and musl link cleanly. Verified locally with `cargo check --target x86_64-unknown-linux-musl`. #### Category 7 — Unresolved merge markers (1 repo) — **DEFERRED, filed hero_slides#71** `hero_slides` `development` HEAD does NOT compile. Commit [`6e39d677`](https://forge.ourworld.tf/lhumina_code/hero_slides/commit/6e39d677) by @despiegk — `chore: auto-commit local changes before pull` (2026-05-19 17:53 +0200) — is a merge that committed **138 unresolved conflict markers across 46 regions in 10 files**, including the binary entry's 42 markers (`main.rs`) and 30 in `generator.rs`. The two parents conflict on incompatible structural intents: - [`45efb36`](https://forge.ourworld.tf/lhumina_code/hero_slides/commit/45efb36) by @Casper — server module-layout refactor (file renames into `jobs/` and `rpc/`) - [`93a3c99`](https://forge.ourworld.tf/lhumina_code/hero_slides/commit/93a3c99) by @despiegk — herolib_ai API migration (ImageRef relocation, error variant renames, agent invocation rewrites) Resolving requires picking the winner per concern; the original authors have the context. Filed [hero_slides#71](https://forge.ourworld.tf/lhumina_code/hero_slides/issues/71) with details + suggested unwinding procedure. The §3 estimate of `≤ 10 min` (based on raw marker count) underestimated by an order of magnitude. Per [`feedback_branch_cleanup_only_own_authored.md`](https://forge.ourworld.tf/lhumina_code/home/raw/branch/main/memory/feedback_branch_cleanup_only_own_authored.md), I left it alone. #### Category 8 — Stale `herolib_core` API (1 repo) — **DIAGNOSIS WAS WRONG** `hero_books`. Phase B confirmed (and local `lab build --release --platforms linux-musl-x86_64 --workspace` confirmed mid-build) that: 1. `herolib_core::base::hero_socket_dir` and `path_var()` ARE exported by current `herolib_core` ([base/mod.rs:8-9 + paths.rs:54,86](https://forge.ourworld.tf/lhumina_code/hero_lib/src/branch/development/crates/core/src/base/mod.rs)). 2. `hero_books` builds cleanly under lab locally — `hero_books_admin` (1.97MB), `hero_books_web` (4.34MB), `hero_books_server` and `hero_do_hero_books` produced before the test timeout. 3. `releases/tag/latest` already has `hero_books_admin`, `hero_books_web`, `hero_do_hero_books` musl-x86_64 assets shipping from a prior workstation upload. The lab-publish.yaml CI fail is not a code issue. Likely: transient infrastructure / asset-upload step / runner pre-condition. Recommend triggering a fresh CI run and harvesting the actual log when the next push lands on hero_books. #### Category 9 — Stale service.toml refs (1 repo) — **DIAGNOSIS WAS WRONG; filed hero_skills#270** `hero_rpc`. `hero_recipes_server` and `hero_recipes_admin` crates DO exist (`example/recipe_server/crates/{hero_recipes_server,hero_recipes_admin}/`). The actual failure surface is a **lab orchestrator bug with nested workspaces**: lab's `executor.rs:54` runs `cargo build -p hero_recipes_server` from `repo_root`, but `repo_root`'s Cargo.toml `exclude`s `example/recipe_server` — so cargo says "package ID specification did not match." The bin is correctly discovered by `discover_binaries_in_repo` (which walks per-workspace cargo metadata), but the executor doesn't carry the bin's workspace root through to `current_dir`. Detailed root-cause analysis and fix shapes at [hero_skills#270](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/270). This blocks any D-07 repo that has nested-workspace examples under `git_root` — hero_rpc is the confirmed instance. ### Net coverage ``` #269 baseline at s127 close: 14 / 31 wired (45%) s128 fixes (CI in flight): +9 ← 5 path-dep + 2 feature + 1 sqlite + 1 rlimit s128 already-shipping: +2 ← hero_code + hero_books (older uploads) Projected after CI runs: 25 / 31 wired (81%) Projected against 35-set: 25 / 35 D-07 (71%) ``` Unblocked remaining (against the 35-set): - 4 special-case deferred (Surface B) — hero_voice, hero_embedder, hero_memory (gnu matrix-swap per D-05), hero_office (private auth), mycelium_network (branch override), hero_os (WASM). Tracked at [#268](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/268)/[#269](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/269) Out of scope sections. - hero_slides — needs [#71](https://forge.ourworld.tf/lhumina_code/hero_slides/issues/71) merge reconciliation. - hero_rpc — needs [#270](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/270) lab patch. ### Followups filed - [hero_slides#71](https://forge.ourworld.tf/lhumina_code/hero_slides/issues/71) — broken merge state on `development`; needs despiegk + Casper. - [hero_skills#270](https://forge.ourworld.tf/lhumina_code/hero_skills/issues/270) — lab executor.rs CWD bug on nested workspaces; blocks hero_rpc CI.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
development
main
integration
feat/openrpc-wasm-dioxus-skills
feature/reenable-blueprint
development_timur
lab-agent-skills-kimi-fixes
latest-dev
lab-latest
latest
Labels
Clear labels   
prio_critical

   
prio_low

   
type_bug

   
type_contact

   
type_issue

   
type_lead

   
type_question

   
type_story

   
type_task
No labels
prio_critical
prio_low
type_bug
type_contact
type_issue
type_lead
type_question
type_story
type_task
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
lhumina_code/hero_skills#269
No description provided.