Skip to content

Reduce avoidable approval prompts from independent shell batches - #1994

Merged
Aaronontheweb merged 5 commits into
devfrom
goal/reduce-fresh-session-approval-spam-live-batching
Aug 18, 2026
Merged

Reduce avoidable approval prompts from independent shell batches#1994
Aaronontheweb merged 5 commits into
devfrom
goal/reduce-fresh-session-approval-spam-live-batching

Conversation

@Aaronontheweb

Copy link
Copy Markdown
Collaborator

Summary

  • freeze a sanitized post-Reduce project scope declaration misses #1990 live sample where four of ten shell calls prompted because independent diagnostics were batched
  • add generic one-operation-per-call guidance across always-loaded rules, shell metadata, and the bundled operations skill
  • strengthen the fresh-session eval so compound commands or unsuccessful shell results cannot pass
  • keep authored compound calls approval-gated and lock those outcomes in the coordinator evidence corpus

This changes model guidance only. It adds no grant, safe verb, shell-authority rule, executable-specific parser, public API, or persistence format.

Measured behavior

Using the same five-run task and primary model configuration:

  • behavior completion: 3/5 to 5/5
  • shell calls: 6 to 5
  • compound shell calls: 1 to 0
  • approval-equivalent events: 1 to 0

Validation

  • full pre-rebase solution run: 7,562 passed; 15 expected skips
  • post-Fix deterministic CI failures across active pull requests #1991 rebase Release build: clean
  • post-rebase focused interaction set: 445 passed; 1 expected Windows-only skip
  • changed-file Slopwatch: 0 findings
  • strict OpenSpec, changed-file formatting, headers, Bash syntax, JSON, and diff checks: pass
  • exact candidate binary swapped locally and daemon health verified

@Aaronontheweb
Aaronontheweb enabled auto-merge (squash) August 18, 2026 19:27
@Aaronontheweb Aaronontheweb added the shell Issues related to the shell tool, since it has the largest security perimeter. label Aug 18, 2026
@Aaronontheweb
Aaronontheweb merged commit c1f17f3 into dev Aug 18, 2026
23 checks passed
@Aaronontheweb
Aaronontheweb deleted the goal/reduce-fresh-session-approval-spam-live-batching branch August 18, 2026 21:35
@Aaronontheweb Aaronontheweb mentioned this pull request Aug 26, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

shell Issues related to the shell tool, since it has the largest security perimeter.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant