22
33This audit records the execution surface the PyAutoBrain Build Agent reasons over
44and ** calls** . The agent must treat every item here as a PyAutoHands capability,
5- not as logic to reimplement inside Brain. Source of truth: the ` autobuild `
6- dispatcher (` PyAutoHands/bin/autobuild ` ) and ` PyAutoHands/CLAUDE.md ` .
5+ not as logic to reimplement inside Brain. Source of truth: the ` autohands `
6+ dispatcher (` PyAutoHands/bin/autohands ` ) and ` PyAutoHands/CLAUDE.md ` .
77
88## Execution capabilities (the Build Agent calls these)
99
10- From ` autobuild help` . ** Mode** is the Build Agent allowlist that * may* route
10+ From ` autohands help` . ** Mode** is the Build Agent allowlist that * may* route
1111the capability (` build.sh ` ); ** Consumer today** is who actually drives it now.
1212The two diverge for the workspace-run surface — see the note below.
1313
@@ -24,26 +24,26 @@ The two diverge for the workspace-run surface — see the note below.
2424| ` script_matrix ` | Output a JSON ` {name, directory} ` matrix for GitHub Actions. | build | ** Heart** ` workspace-validation.yml ` (CI) |
2525| ` aggregate_results ` | Aggregate per-job JSON into a release-readiness report. | build, release | ** Heart** ` workspace-validation.yml ` (CI) |
2626| ` slow_skip_check ` | Surface SLOW / NEEDS_FIX entries in workspace ` no_run.yaml ` . | build | ` run_all.py ` + Heart ` test_run ` check |
27- | ` repro_command ` | Emit the shell command autobuild uses to run one script. | build | triage (manual handoff) |
27+ | ` repro_command ` | Emit the shell command autohands uses to run one script. | build | triage (manual handoff) |
2828| ` bump_colab_urls ` | Rewrite Colab URLs in cwd from an old to a new date-tag. | build, deploy | ` release.yml ` (CI) + deploy |
2929
3030** Reading the Consumer column — the run\* seam is latent, not active.** The
31- capabilities still * exist* and match the ` autobuild ` dispatcher and the
31+ capabilities still * exist* and match the ` autohands ` dispatcher and the
3232` build.sh ` allowlists exactly; the * Mode* column remains accurate as the
3333allowlist of record. What changed is the driver. The workspace full-run →
3434report → issue flow moved ** out of ` release.yml ` into PyAutoHeart's
3535` workspace-validation.yml ` ** (see ` PyAutoHands/docs/internals.md ` and the
3636` release.yml ` comment recording the removed ` run_scripts ` / ` run_notebooks ` /
3737` analyze_results ` jobs). Heart checks these primitives out from Build and reuses
3838them directly — calling ` run.py ` / ` run_python.py ` / ` script_matrix.py ` /
39- ` aggregate_results.py ` , ** not** ` autobuild run_*` and ** not** `pyauto-brain build
39+ ` aggregate_results.py ` , ** not** ` autohands run_*` and ** not** `pyauto-brain build
4040run_ * ` . Nothing invokes ` run` / ` run_python` / ` run_all` through the Build Agent
4141today, so their build-mode allowlist entries are a ** latent seam** — kept so the
4242Build Agent could re-own execution without churn, but ** Heart is the live
4343consumer** . This is the execution/health boundary the audit below argues for,
4444now made visible in the table itself.
4545
46- Underlying implementation assets the agent never re-owns: ` autobuild /run_python.py` ,
46+ Underlying implementation assets the agent never re-owns: ` autohands /run_python.py` ,
4747` run.py ` , ` generate.py ` , ` script_matrix.py ` , ` aggregate_results.py ` ,
4848` tag_and_merge.sh ` , ` build_util.py ` , the ` release.yml ` GitHub workflow, and the
4949per-workspace ` config/build/{no_run,env_vars,visualise_notebooks}.yaml ` .
0 commit comments