Official Codebase of 3DCodeVerse
Python
docs: RUNBOOK no longer says gemini-cli transcripts feed repair-pair mining; SKILLS_LEDGER dates its status vocabulary
gaoypengf0e4b053eb63
diff --git a/harness/docs/RUNBOOK.md b/harness/docs/RUNBOOK.mdindex c8acb2d7..16fa642c 100644--- a/harness/docs/RUNBOOK.md+++ b/harness/docs/RUNBOOK.md@@ -63,7 +63,7 @@ round, then ≈ $0.36 / ~7 min per refine (give scenes `--max-minutes 60`).| id | what runs | notes ||---|---|---|-| `gemini-cli:gemini-3.7-flash` (default) | `gemini -m … --approval-mode yolo --skip-trust --output-format json`, prompt on stdin; every spatial tool over MCP | cheapest agentic path; transcripts feed repair-pair mining; see gotchas below |+| `gemini-cli:gemini-3.7-flash` (default) | `gemini -m … --approval-mode yolo --skip-trust --output-format json`, prompt on stdin; every spatial tool over MCP | cheapest agentic path; see gotchas below || `single-shot:gemini:gemini-3.7-flash` | one structured-output call → multi-file envelope, no tools | fastest/cheapest; baseline for "raw model" deltas || `claude-code:<model>` | `claude -p --dangerously-skip-permissions --setting-sources project --settings '{"skillOverrides":…}' --mcp-config trajectories/<label>_rNN/mcp.json --strict-mcp-config …`, prompt on stdin, `CLAUDE_CODE_DISABLE_BUNDLED_SKILLS=1` | local subscription — test lightly; no user setting source, so `~/.claude/settings.json` (effortLevel, hooks, env) does not reach a session || `codex:<model>[@<effort>]` | `codex exec --json -C ws --sandbox workspace-write -c model_reasoning_effort=high -c skills.bundled.enabled=false -c mcp_servers.3dcode.… -`, prompt on stdin | subscription; MCP tools need `default_tools_approval_mode="approve"` (harness passes it); reasoning effort is always stated (`Settings.agents.codex_reasoning_effort`, default `high`; `codex:gpt-5.6-sol@medium` per id, `""` to defer to `~/.codex/config.toml`) |diff --git a/harness/docs/SKILLS_LEDGER.md b/harness/docs/SKILLS_LEDGER.mdindex b46dfab4..11239059 100644--- a/harness/docs/SKILLS_LEDGER.md+++ b/harness/docs/SKILLS_LEDGER.md@@ -316,7 +316,8 @@ it lands these four rows stay `mixed`, not `measured`.it (`BuildResult.gates`, appended by the round built or not); runs before that date stilldo not.* **A run is a sample iff it has a round carrying a gate report.** `RunRecord.status` says- how the round loop *ended* (`passed`/`plateau`/`budget`/`failed`), not whether gates ran.+ how the round loop *ended* (then `passed`/`plateau`/`budget`/`failed`; since 2026-09-22+ only the stop reason, `max_rounds`/`budget`/…), not whether gates ran.Filtering on status dropped every `budget` run — about half the corpus — and would haveadmitted a `passed` run with no gated round as a clean 0.* **A budget cap tuned for the judged score starves the gate readout.** The `bbox-contract`