feat(agentic): fresh-interpreter + preinstalled-packages guidance for code agents#17
Merged
Merged
Conversation
…ckage guidance The stateless code-analyst path renders `render_code_exec_supplement`, which never told the model that each `run_code` call is a fresh E2B interpreter or which packages are available. Traces showed the model losing variables between calls (NameError on `df`) and importing unavailable packages (`arch`). Extend the shared workstyle supplement with two principled additions: - Fresh interpreter per execution: no state survives between `run_code` calls, so every script must be fully self-contained (reinforces batching guidance). - Available packages: render `domain.code_exec_preinstalled` with an explicit "other packages are NOT installed" fence. Enrich the preinstalled list to reflect the real E2B template (`aieng-forecasting[agentic,numerical,llm]` on e2bdev/code-interpreter): add scipy (base image), statsforecast/darts/lightgbm (numerical extra) so the list steers the model toward genuinely-available tools. Applied to the default field value and SP500/TSX domains; oil left untouched (it does not call the supplement). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Instruction-level half of the sandbox findings (persistence itself is Phase-2): the code-exec workstyle supplement now states every run_code call is a brand-new interpreter (scripts must be self-contained) and renders the verified preinstalled-package list (enriched to include darts/lightgbm/statsforecast/scipy — all confirmed present in the E2B template; 'arch' correctly absent). Oil/bootcamp paths untouched. 684 passed; pre-commit clean.
🤖 Generated with Claude Code