ai-wiki-toolkit-linux-arm64 0.1.25 → 0.1.27
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/README.md +23 -2
- package/bin/aiwiki-toolkit +0 -0
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
# ai-wiki-toolkit-linux-arm64
|
|
2
2
|
|
|
3
3
|
This package contains the `aiwiki-toolkit` executable for `linux-arm64-glibc`.
|
|
4
|
-
It is published as the platform-specific binary package for `ai-wiki-toolkit` `0.1.
|
|
4
|
+
It is published as the platform-specific binary package for `ai-wiki-toolkit` `0.1.27`.
|
|
5
5
|
Most users should install `ai-wiki-toolkit` instead of using this package directly.
|
|
6
6
|
|
|
7
7
|
---
|
|
@@ -91,6 +91,11 @@ That is why the installer manages both wiki files and prompt wiring together:
|
|
|
91
91
|
- `.agents/skills/ai-wiki-reuse-check/` and `.agents/skills/ai-wiki-update-check/` provide repeatable end-of-task checks for Codex-style agent runs
|
|
92
92
|
- `.agents/skills/ai-wiki-clarify-before-code/` and `.agents/skills/ai-wiki-capture-review-learning/` help agents clarify ambiguous requests and preserve reusable review feedback
|
|
93
93
|
|
|
94
|
+
Some runtimes can discover repo-local `.agents/skills/` files on disk but still not expose those
|
|
95
|
+
skills to the active model session. The managed `ai-wiki/_toolkit/system.md` includes fallback
|
|
96
|
+
paths so agents can manually read the relevant `SKILL.md` and `references/` files when runtime skill
|
|
97
|
+
exposure is missing.
|
|
98
|
+
|
|
94
99
|
The toolkit does not replace coding agents. It gives them a shared repo-local memory layer so they can avoid repeating the same review issues, misunderstanding the same requirements, or rediscovering the same fixes.
|
|
95
100
|
|
|
96
101
|
## Task-Aware Context Routing
|
|
@@ -254,6 +259,7 @@ npm install -g ai-wiki-toolkit@latest
|
|
|
254
259
|
after install. That is intentional: they are placeholders for rules and decisions your team has
|
|
255
260
|
actually made, not generic package defaults.
|
|
256
261
|
- `ai-wiki/_toolkit/system.md` is the managed entrypoint for package-managed repo guidance and evolving read order.
|
|
262
|
+
- `ai-wiki/_toolkit/system.md` also includes runtime skill fallback guidance for agents whose active session does not expose repo-local `.agents/skills/`.
|
|
257
263
|
- `ai-wiki/index.md` is a repo-owned map, not a package upgrade surface.
|
|
258
264
|
- `ai-wiki/_toolkit/workflows.md` carries the managed baseline workflows that package upgrades can refresh.
|
|
259
265
|
- Agents should extend user-owned workflow docs instead of editing `_toolkit/**`.
|
|
@@ -427,6 +433,21 @@ aiwiki-toolkit consolidate queue --since 14d --handle your-handle
|
|
|
427
433
|
|
|
428
434
|
This writes regenerated local reports under `ai-wiki/_toolkit/consolidation/` and prints the queue to stdout. The queue suggests one action per draft cluster: keep, refine, promotion candidate, conflict, or supersession. It does not edit user-owned AI wiki docs or create shared conventions, review patterns, problems, features, or decisions; those still require human confirmation.
|
|
429
435
|
|
|
436
|
+
To summarize first-attempt product impact from a captured eval run:
|
|
437
|
+
|
|
438
|
+
```bash
|
|
439
|
+
aiwiki-toolkit eval impact report --run-dir /path/to/eval-run
|
|
440
|
+
aiwiki-toolkit eval impact report --run-dir /path/to/eval-run --format json
|
|
441
|
+
```
|
|
442
|
+
|
|
443
|
+
This reads an existing run directory with `metadata.json`, result captures, optional
|
|
444
|
+
`score.json` files, and optional `confounds.json`. It compares the run's primary variants,
|
|
445
|
+
normally `no_aiwiki_workflow` versus `aiwiki_ambient_memory_workflow`, using first-attempt
|
|
446
|
+
metrics only: `first_pass` captures count toward the signal, while `final` repair captures
|
|
447
|
+
stay diagnostic. The command reports first-attempt success rate, average score, attempts, human
|
|
448
|
+
nudges, changed files, untracked files, and whether the run is ready for shareable causal claims.
|
|
449
|
+
It does not run agents or mutate eval artifacts.
|
|
450
|
+
|
|
430
451
|
To diagnose missing starter pointers, stale managed prompt blocks, or rule drift and print copy-paste upgrade starters:
|
|
431
452
|
|
|
432
453
|
```bash
|
|
@@ -445,7 +466,7 @@ This command does not rewrite user-owned repo docs. It prints which paths need a
|
|
|
445
466
|
- `ai-wiki/people/<handle>/index.md`
|
|
446
467
|
- `ai-wiki/metrics/index.md`
|
|
447
468
|
|
|
448
|
-
It also checks whether the managed `.gitignore` block is present
|
|
469
|
+
It also checks whether the managed `.gitignore` block is present, whether local identity, telemetry, or generated-view paths are still tracked in the git index from older versions, and whether managed system rules include the runtime skill fallback used when repo-local AI wiki skills are present but not exposed by the active agent runtime. If local-state paths are still tracked, `doctor` prints a one-time `git rm --cached` command to untrack them.
|
|
449
470
|
|
|
450
471
|
To remove the managed layer while keeping your user-owned wiki documents:
|
|
451
472
|
|
package/bin/aiwiki-toolkit
CHANGED
|
Binary file
|
package/package.json
CHANGED