@rackbops/ac-agent 2.0.0-alpha.8 → 2.0.0-alpha.9

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/data/README.md ADDED
@@ -0,0 +1,40 @@
1
+ # `packages/agent/data/`
2
+
3
+ Vendored, unmodified copies of source files owned by other repos, embedded into the agent's bundle
4
+ as text (see `../text-plugin.mjs`).
5
+
6
+ ## `Get-ToolchainInventory.ps1`
7
+
8
+ A byte-identical copy of `artifact-console/data/Get-ToolchainInventory.ps1` from
9
+ [`Rackbops/artifacts-console`](https://github.com/Rackbops/artifacts-console) (the 1.x app), pinned
10
+ to commit `7280ae6`:
11
+
12
+ - 1,397 lines, 77,956 bytes
13
+ - LF line endings, no BOM, ASCII-only
14
+ - sha256 `28db241b5aa42202c329c115a0de5a28049b04a421d980f9ec463e109e9d0912`
15
+
16
+ The hash is platform-stable because both repos' `.gitattributes` normalize this file to
17
+ `text=auto eol=lf`, and `toolchain-inventory.test.ts` asserts the embedded text hashes to the value
18
+ above -- a re-vendored script whose bytes moved fails that test rather than silently drifting.
19
+
20
+ **Re-copy, never edit.** The script is shipped and run as-is (`packages/agent/src/scans/
21
+ toolchain-inventory.ts`); a local edit here would break the byte-compare against 1.x's own output,
22
+ which is the entire point of vendoring rather than rewriting it. To pick up a newer 1.x revision:
23
+ copy the file from the new commit, update the commit hash and the three numbers above, and re-run
24
+ the embedding/drift-guard tests (`toolchain-inventory.test.ts`) -- they also pin the script's
25
+ catalog names and schema-version literal, so a re-vendor whose shape moved is caught there, not
26
+ discovered on a card.
27
+
28
+ **Provenance note (52b, landed 2026-09-13):** `Rackbops/artifacts-console`'s `main` had moved to
29
+ `1e7cfe5` by the time this landed -- 424 lines away from the `7280ae6` pin (`git diff --stat 7280ae6
30
+ 1e7cfe5 -- artifact-console/data/Get-ToolchainInventory.ps1`: 414 insertions, 10 deletions). The
31
+ difference is one self-contained feature, a tools-site page-data push (issues #636/#637: seven new
32
+ functions -- `Resolve-ToolchainScpExe`, `Get-ToolchainSidecarFiles`, `New-ToolchainPageManifest`,
33
+ `Resolve-ToolchainGhExe`, `New-ToolchainFiledSnapshot`, `Invoke-ToolchainPageDataPush`,
34
+ `Test-ToolchainPageDataChanged` -- plus a new optional `-PageDataTarget` parameter and a
35
+ null-guard around `$env:SystemRoot` in `Resolve-ToolchainSshExe`) -- none of it touches
36
+ `Publish-ToolchainReport` (the report + sidecar writer) or the local `-OutFile -NoRender` path this
37
+ agent invokes, and the new main-flow block only runs when `-PageDataTarget` is set, which the agent
38
+ never passes. So `7280ae6` and `1e7cfe5` are behaviorally identical for this agent's invocation; the
39
+ pin was kept rather than churned. Re-pinning to a newer 1.x commit is a fleet-pass decision, not
40
+ something this PR does on its own.