# PyAutoBrain

## Agents & workflows

### Conductors — plan and coordinate
- **intake** — File classified prompts; never start development · Use the intake skill.
- **batch** — Propose a review-budgeted batch; collect its results · Use the batch skill.
- **community** — Triage community threads; human-approved replies and development handoffs · Use the community skill.
- **feature** — Select, size and plan features · Use the feature skill.
- **bug** — Classify regressions and plan repairs · Use the bug skill.
- **refactor** — Plan behavior-preserving restructuring · Use the refactor skill.
- **workspace** — Plan or survey examples and tutorials · Use the workspace skill.
- **eyes** — Survey figures and route accepted critiques · Use the eyes skill.
- **profiling** — Plan campaigns; ingest results; triage profiling drift · Use the profiling skill.
- **cortex** — Pull science runs and record human checkins · Use the cortex skill.
- **hygiene** — Prioritize upkeep and delegate fixes · Use the hygiene skill.
- **clone** — Plan assistant births or synchronize generic files · Use the clone skill.
- **build** — Consult vitals and coordinate Hands execution · Use the build skill.
- **release** — Coordinate release rehearsal, validation and nightly execution · Use the release skill.
- **health** — Assess health and route repairs · Use the health skill.

### Faculties — read-only advice
- **vitals** — Read authoritative Heart readiness · Use the vitals skill.
- **review** — Prepare independent CLEAN/FINDINGS/BLOCKED branch review · Use the review skill.
- **memory** — Recall cited scientific and task history · Use the memory skill.
- **samplers** — Inspect sampler coverage, tiers and benchmarks · Use the samplers skill.
- **sizing** — Estimate task difficulty · Use the sizing skill.

### Workflows — carry out a procedure
- **brain** — Run a named PyAutoBrain conductor or faculty directly through bin/pyauto-brain. · Use the brain skill.
- **ci-speedup** — Find the slowest CI parts on the PyAutoHeart board (smoke scripts, unit tests, workflow gates), and drive speed-ups through the dev flow. · Use the ci-speedup skill.
- **docs** — Start PyAutoLabs documentation, tutorial, notebook, example, or narrative-prose work through the Brain development workflow with the docs wo · Use the docs skill.
- **feedback** — Draft user-reviewed feedback on using PyAutoLabs software and assistants (goals, successes, friction, evidence). · Use the feedback skill.
- **issue-cleanup** — Reconcile the PyAuto GitHub issue trackers — audit open issues against Mind completion records and merged PRs, bucket them, close only what  · Use the issue-cleanup skill.
- **prm** — Judge every CI run and matrix leg, merge human-authorized PRs when green, then close the issue and Mind task and clean the worktree. · Use the prm skill.
- **repo-cleanup** — Periodic hygiene sweep across PyAuto repos and worktrees — audit stale feature branches, tracking refs, stashes, and dirty checkouts, then e · Use the repo-cleanup skill.
- **research** — Start exploratory scientific, algorithmic, architectural, or prior-art investigation through the PyAutoBrain development workflow with the r · Use the research skill.
- **route** — Infer the appropriate PyAutoBrain command or development work type from a free-form PyAutoLabs request. · Use the route skill.
- **sampler-pipeline** — Trial a new non-linear sampler via ingest → prototype → profile → promote, up to a full PyAutoFit implementation. · Use the sampler-pipeline skill.
- **ship-library** — Validate and ship PyAuto library changes to pending-release PRs, including downstream impact and Mind state. · Use the ship-library skill.
- **ship-workspace** — Validate and ship PyAuto workspace or tutorial changes behind the library-first gate, then update Mind state. · Use the ship-workspace skill.
- **start-dev** — Plan and register PyAutoLabs development work, then route it to a library or workspace worktree. · Use the start-dev skill.
- **start-dev-for-user** — Start a PyAutoLabs development task for another contributor, including plan, issue, branch, and handoff preparation without claiming the loc · Use the start-dev-for-user skill.
- **start-library** — Create or resume claimed PyAuto library worktrees after start-dev and prepare them for source edits. · Use the start-library skill.
- **start-workspace** — Set up PyAutoLabs workspace or tutorial development — attach or create workspace worktrees, inspect upstream library API changes, and regist · Use the start-workspace skill.
- **update-issue** — Post a progress update to a GitHub issue from the current CLI session — commits, summary, and remaining work. · Use the update-issue skill.

## Review overnight work
⏸ PyAutoLabs/PyAutoBrain/nightly-release.yml — blocked at a gate, no change made (5h)
✗ PyAutoLabs/PyAutoHands/python_matrix.yml — cancelled (3d)
  Use the bug skill. overnight: PyAutoLabs/PyAutoHands/python_matrix.yml concluded cancelled — https://github.com/PyAutoLabs/PyAutoHands/actions/runs/37294378873

## Maintenance
- Investigate slow tests / CI: Use the ci-speedup skill. Read Heart's current performance evidence, identify the largest actionable bottleneck and plan a targeted fix.
- Review code quality: Use the hygiene skill.
- Reconcile issue trackers: Use the issue-cleanup skill.
- Clean up branches and worktrees: Use the repo-cleanup skill.
Performance evidence: https://pyautolabs.github.io/PyAutoHeart/
- docstrings: 28 adjacent documentation boundaries in 19 files across 3/5 repos; 0 parse errors — Use the refactor skill.
- escapes: 1 script(s) across 5 repo(s) with LaTeX damaged by string escapes: 2 warned, 1 silent (0 file(s) have ONLY silent damage, which a warning-only sweep would miss); 0 read error(s) — Use the refactor skill.
- refs: 60 folder-list defects (60 dead references, 0 undocumented folders) in 24 files across 3/5 repos; 251 unresolvable refs skipped — Use the refactor skill.
- optdeps: 1 smoke-listed script uses a gated API with no optional-dependency skip guard (nufftax:1) — fails the CI matrices that omit the extras — Use the refactor skill.
- deps: 38 capped/pinned specifiers across 10 pyproject.toml (20/20 code repos present) — Use the dep-audit skill.
- docs: 33 api .rst files, 82 currentmodule directives across 4 repos with docs/api (20/20 code repos present) — Use the audit-docs skill.
- config: 25 library config keys absent downstream (review/mirror): autofit_workspace:5 autogalaxy_workspace:19 autolens_workspace:1 — Use the refactor skill.