@williambeto/ai-workflow 2.7.1 → 2.8.1

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/CHANGELOG.md CHANGED
@@ -1,5 +1,74 @@
1
1
  ## [unreleased]
2
2
 
3
+ ## [2.8.1] - 2026-07-14
4
+
5
+ ### 🐛 Bug Fixes
6
+
7
+ - Route natural workspace mutations to Astra instead of stopping after a
8
+ specialist prints code
9
+ - Keep Phoenix limited to bounded remediation of observed findings
10
+ - Require execution-scoped repository changes and proportional validation for
11
+ successful workspace delivery
12
+ - Preserve explicit read-only, preview, already-satisfied, clarification, and
13
+ non-delivery outcomes without manufacturing changes
14
+ - Prevent natural phrases such as `next step` from being interpreted as a
15
+ Next.js stack request
16
+ - Propagate observed OpenCode commands to the evidence ledger and derive test
17
+ actions from the delivered diff
18
+ - Prevent duplicate finalization and require exactly one observed Astra
19
+ delegation in the OpenCode fallback path
20
+ - Route mixed inspection-and-change requests to Astra whenever the request has
21
+ write intent
22
+ - Allow read-only Git, `sed`, and `awk` inspection while continuing to block
23
+ mutations, and exclude ignored build artifacts from delivery deltas
24
+ - Keep validation proportional for documentation, simple code, backend scripts,
25
+ and React applications while retaining blocking build requirements for UI
26
+
27
+ ### 🧪 Testing
28
+
29
+ - Cover routing ownership, delivery outcomes, answer-only confinement, project
30
+ context compatibility, execution delta attribution, and generated OpenCode
31
+ permissions through reusable invariants
32
+ - Keep pre-existing dirty work from satisfying delivery or test-action evidence
33
+ - Validate the corrected workflow with a real GPT-5.6 Sol OpenCode delivery and
34
+ record GPT-5.3 Codex Spark medium as experimental after observed incomplete
35
+ delivery was correctly blocked
36
+
37
+ ### 📚 Documentation
38
+
39
+ - Align Atlas, quickstart, ownership, runtime-flow, evidence, and skill
40
+ governance documentation with the natural-request delivery contract
41
+ - Document the distinction between CLI-enforced gates and model-qualified
42
+ direct Atlas sessions
43
+
44
+ ## [2.8.0] - 2026-07-14
45
+
46
+ ### 🚀 Features
47
+
48
+ - Add governed skill scaffolding, frontmatter linting, and contract tests
49
+ - Introduce executable DEEP SDD checkpoints with hash-bound approval receipts
50
+ - Add Orion technical-plan validation and Astra file-scope enforcement
51
+ - Confirm high-risk OpenCode actor identity through machine-readable runtime evidence
52
+
53
+ ### 🐛 Bug Fixes
54
+
55
+ - Preserve low- and medium-risk OpenCode graceful degradation behavior
56
+ - Resolve packaged evidence schemas from the installed package root
57
+ - Prevent metadata sentinels from being treated as Astra file mutations
58
+
59
+ ### 🧪 Testing
60
+
61
+ - Cover external, missing, incompatible, duplicate, and concurrent workflow ledgers
62
+ - Add real OpenCode consumer smoke coverage and actor-confirmation adapter tests
63
+ - Enforce Node 20 consumer compatibility and Node 22 contributor validation
64
+
65
+ ### ⚙️ Miscellaneous Tasks
66
+
67
+ - Add sequential release verification across build, tests, validation, packaging, and runtime smoke
68
+ - Align CI quality gates with the supported Node toolchain
69
+
70
+ ## [2.7.1] - 2026-07-07
71
+
3
72
  ### 🚀 Features
4
73
 
5
74
  - *(adapters)* Harden Codex adapter with robust system boundaries
package/README.md CHANGED
@@ -148,6 +148,7 @@ Agents are installed as prompts in your `.ai-workflow/agents/` directory. You ca
148
148
  | `npx aw collect-evidence` | Collect and compile delivery evidence to `EVIDENCE.json` |
149
149
  | `npx aw doctor` | Diagnose installation health |
150
150
  | `npx aw clean` | Remove all AIWK configuration from the project |
151
+ | `npx aw skill create <name>` | Scaffold a workspace-local skill with valid frontmatter |
151
152
 
152
153
  ---
153
154
 
@@ -200,7 +201,7 @@ Everything is plain text — readable, auditable, and version-controlled.
200
201
  Beyond the 6 core agents, AIWK ships 25 domain skills that agents activate as context lenses when the request profile warrants it. You can also author your own:
201
202
 
202
203
  ```
203
- .ai-workflow/skills/
204
+ .agents/skills/
204
205
  ├── architecture/
205
206
  ├── backend-development/
206
207
  ├── cyber-security/ # activated by Sage on audit tasks
@@ -246,7 +247,7 @@ Run the full validation suite before submitting a pull request:
246
247
  npm run validate
247
248
  ```
248
249
 
249
- All 22 quality gates must pass.
250
+ All 23 quality gates must pass.
250
251
 
251
252
  ---
252
253