feat(nano): care / early-delivery / verification guidelines (PR 10) - #889
Merged
Conversation
Restores, as one line each, the general practices whose absence the first TB 2.1 nano run exposed (analysis in my-docs/clawcodex-nano/): * copy-before-probing for state a task needs intact — db-wal-recovery was lost because the first exploratory sqlite3 open of the database reset its invalid WAL, destroying what the task asked to recover * minimal-working-deliverable-early — four tasks died at the per-task wall clock mid-polish with no passing artifact in place, on tasks the same model solved in pi's run by crossing the line earlier * verify-requirements-before-finishing — three tasks ended with a confident summary the verifier rejected All three are distillations of clawcodex's full-prompt sections (Executing actions with care / Doing tasks), not benchmark-specific knowledge. Fixed payload grows ~120 est. tokens (~2.4K total, budget 2.5K); 59 nano tests pass. 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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Three one-line general-practice guidelines restored to the nano prompt, each mapping to an observed TB 2.1 failure class (destructive first probe on db-wal-recovery; wall-clock deaths with no artifact in place; confident-wrong finishes). Distilled from clawcodex's full-prompt sections — no benchmark-specific knowledge. Payload +~120 est. tok (≈2.4K, budget 2.5K). 59 nano tests pass.
🤖 Generated with Claude Code