issues Search Results · language:Dune language:Python language:JavaScript language:JavaScript language:JavaScript linked:pr
Filter by
4.5M results
Problem
SESSION_PROMPT.md step 4 (the Phase A silent-rot audit) has drifted from the actual scripts/audit_phase_a.py
implementation in two ways, and the lock test tests/test_session_prompt_phase_a_audit.py ...
priority:high
jt-mchorse
Roadmap rank 8. evals.py recomputes from scratch with no memory. Add a named baseline + regression diff so the suite
becomes the CI verification gate AGENTS.md describes, catching silent routing/policy ...
enhancement
oldreceipt
Roadmap rank 5. No safe way today to preview a ToolPolicy change. Replay past runs / the eval corpus against a candidate
policy and show allow - deny flips. Makes the policy layer editable with confidence ...
enhancement
Roadmap rank 6. The backend already emits rich SSE governance/runner events that App.tsx currently discards. Render a
live runner timeline during a response and add an AbortController-backed Stop button. ...
enhancement
Roadmap rank 4. All current eval cases are happy-path. Add an adversarial set asserting that injection/escalation
messages yield NO high-risk tool calls. Turns the eval+policy pillars into a demonstrable ...
enhancement
Roadmap rank 9. Infinite handoff loops are the #1 multi-agent failure mode and there is no cap today. Add a
deterministic step/transition budget in run_orchestrator that emits a budget_exhausted trace ...
enhancement
ready-for-agent
Roadmap rank 1. /diagnostics reports ok even when Ollama is unreachable or the configured model is not pulled, so
failures only surface as cryptic chat-time runner errors. Add an active check (reuse the ...
enhancement
ready-for-agent
Roadmap rank 2. The just-merged persistent-conversations path builds team input via conversation_context() with NO size
bound — long chats will overflow the local Ollama model. Mirror the existing CONTEXT_BUNDLE_MAX_CHARS ...
bug
ready-for-agent
Roadmap rank 3 (CRITICAL). Approvals are currently created from keyword-PREDICTED tools, so any tool the model actually
selects at runtime executes WITHOUT going through the approval gate — a silent policy ...
bug
ready-for-agent
Contexto — Causa Raiz do ISE pSEO (2026-06-18)
Problema: O CI gate .github/workflows/audit-seo-notfound.yml usa grep para detectar notFound() em 16 prefixos de rota
protegidos. O grep atual NÃO detecta ...
bug
infrastructure
P0
tjsasakifln
Learn how you can use GitHub Issues to plan and track your work.
Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub IssuesProTip! Restrict your search to the title by using the in:title qualifier.
Learn how you can use GitHub Issues to plan and track your work.
Save views for sprints, backlogs, teams, or releases. Rank, sort, and filter issues to suit the occasion. The possibilities are endless.Learn more about GitHub IssuesProTip! Restrict your search to the title by using the in:title qualifier.