Skip to content

feat: add Lisa-backed StarIntel expert shell - #215

Draft
lost-rob0t wants to merge 26 commits into
masterfrom
feat/lisa-expert-shell
Draft

lost-rob0t wants to merge 26 commits into
masterfrom
feat/lisa-expert-shell

Conversation

@lost-rob0t

@lost-rob0t lost-rob0t commented Sep 17, 2026 •

Copy link
Copy Markdown
Owner

Summary

Adds a Common Lisp StarIntel operator shell backed by a per-session Lisa inference engine.

Expert behavior

  • parses operator input into shell-request facts
  • Lisa rules infer shell-plan facts with operation, risk, rule name, and reason
  • read plans execute immediately
  • write/destructive plans are blocked unless explicitly confirmed with --yes
  • why exposes the selected rule, risk class, reason, and trace
  • Lisp-form commands are read as data with *READ-EVAL* disabled; no arbitrary user-input eval

Operations

  • health/status and server info
  • document get/search/submit/delete
  • target list/get/create, including transient targets
  • dataset size
  • message groups, user/platform/group queries
  • social posts by user
  • raw GET/POST/PUT/DELETE through the existing StarIntel client boundary

Packaging

  • new starintel-expert-shell ASDF program system (star-expert)
  • Lisa added to Qlot manifest/lock
  • standalone starintel-expert-shell-tests ASDF suite
  • dedicated Expert Shell GitHub Actions workflow runs the Lisa suite and builds the executable
  • operator documentation in doc/expert-shell.org

Safety / boundaries

The shell is not a privileged back door: all execution stays behind star.api.client, so server auth, tenant isolation, quota, and validation remain authoritative. Mutating-operation confirmation is implemented as Lisa rules, and tests assert that an unconfirmed mutation never reaches transport.

Test command

qlot install
qlot exec sbcl --non-interactive --no-userinit --no-sysinit \
  --eval '(require :asdf)' \
  --eval '(asdf:test-system :starintel-expert-shell-tests)'

@lost-rob0t
lost-rob0t marked this pull request as draft September 17, 2026 04:06

Copy link
Copy Markdown
Owner Author

PR-recovery gate correction: converted this PR to draft. The implementation may be useful, but the current PR does not identify an owning open issue and I found no matching expert shell issue in the complete server issue search. Under current AGENTS.md, production/repository mutation is issue-first and requires the real five-voter/YAGNI preflight before implementation authority; no such evidence is recorded here.

Preserve this branch as implementation evidence at exact head 45bbd72c7ec3b5057a901f300fd649e3c672c295. Before returning it to ready-for-review, attach/reconcile it to the canonical owning issue, record actual voter dispositions (not reconstructed votes), re-check its API/client assumptions against current /v1 authority, and require exact-head repository gates in addition to the dedicated Expert Shell workflow. No code was changed by this recovery pass.

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