Skip to content

feat(skills): make native specialist workflows primary - #5

Merged
decebal merged 1 commit into
mainfrom
feat/refactor-workflow
Oct 25, 2026
Merged

decebal merged 1 commit into
mainfrom
feat/refactor-workflow

Conversation

@decebal

@decebal decebal commented Oct 25, 2026 •

Copy link
Copy Markdown
Contributor

Allstack now exposes native product and engineering skills as its main interface. Builders can plan or review independently, then enroll accepted work into durable delivery when handoff or shipping is needed.

  • Package seven original methods for Codex and Claude: shape, plan, review, QA, ship, recall, and handoff. Skills-only setup needs no delivery run, Chronis initialization, or Rust project profile.
  • Keep AllSource history, Chronis claims, Prime provenance, immutable handoff, and exact-commit publication underneath. Add standalone source-linked knowledge, active-job discovery, and adoption of existing native work without repeating model execution.
  • Deliver Spendcheck category filtering through real Claude planning, Codex implementation, native recall, and Claude review/QA. Preserve the frozen acceptance inputs and existing unfiltered behavior.

Verification

  • Validate all seven skill files.
  • Install both host packages and verify idempotence without creating task/run stores.
  • Upgrade an actual previous installation while denying source-checkout reads; preserve fifteen existing resources and policy.
  • Real Claude native shaping/planning; real Codex application implementation and standalone Prime recall; real Claude native review/QA. No model overrides or permission bypasses.
  • Before application change: two baseline tests passed; four category cases failed. After change: all eight unchanged application tests passed.
  • Inspect actual tool calls, three-file model diff, frozen source hashes, and four independently repeated CLI outcomes.
  • Real Git/Chronis staged-deletion qualification: old native CLI failed, repaired CLI verified one commit; only remote discovery reads were mocked.
  • Focused process owner-loss assertion passed twenty separate executions after PID readiness repair; original deadlines and reaping checks retained.

Design: native specialist realignment. Evidence: installed skills and application verification.

Local scope: sequential same-machine hosts, lexical project Prime, verified commit and open PR. Merge, deployment, and browser QA remain outside this proof.

Exact verified commit: 1093bd3. Managed AllSource run: 9dc4c81f-73ca-4b92-89a2-be3c2ca44deb.

  • Final commit: 53 runtime/native tests, eight application tests, four host-fixture tests, formatting, lint, and source checks passed. Two real-Chronis qualifications remain explicitly skipped in the portable suite; staged-deletion qualification passed separately.
  • Exact remote branch and PR head read back; Chronis task completion and Prime outcome confirmed.

CI: Quality on Linux and macOS passed all three checks. Each platform executed 53 runtime tests (2 explicit skips), 8 Spendcheck application tests, and 4 handoff fixture tests.

Expose independent product and engineering skills while retaining durable coordination behind enrollment, handoff, and publishing. Add source-linked standalone Prime knowledge and verify an expense category change through authenticated Claude and Codex sessions. Preserve legacy resources and event history, including already staged deletions. Keep process ownership assertions reliable during PID publication.

Allstack-Run: 9dc4c81f-73ca-4b92-89a2-be3c2ca44deb
@decebal
decebal force-pushed the feat/refactor-workflow branch from 9aaf7c4 to 1093bd3 Compare October 25, 2026 21:11
@decebal decebal changed the title docs(workflows): propose specialist-first product realignment feat(skills): make native specialist workflows primary Oct 25, 2026
@decebal
decebal marked this pull request as ready for review October 25, 2026 21:22
@decebal
decebal merged commit 7c6f580 into main Oct 25, 2026
3 checks passed
@decebal
decebal deleted the feat/refactor-workflow branch October 25, 2026 21:28
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant