Intake organ Memory gate, board phase 2, RTD organism currency

ORGANISM.md

@@ -22,6 +22,7 @@ organs plan, build, test and release it, and you make every judgment call.
 | **Hands** | PyAutoHands | Builds and releases — packaging, tagging, notebook generation, PyPI via `release.yml`. A pure executor: runs no readiness checks and never re-derives a gate decision. |
 | **Pulse** | PyAutoPulse | Feels *how fast the organism runs* across the `<lib>_profiling` project repos. Owns profiling campaign intent and pending domain tasks, the profiling instance registry, the versioned `profiling-summary` read contract, the ingest receipts (resolved commit per project per render) and the cross-project board. Validates the exchange contract only; project repos own producers, results, pins, drift policy and their own Pages page. Never moves pins, combines unmatched timings, applies the compile threshold to runtime, computes an ecosystem-wide speed score or issues a readiness verdict; the Brain's profiling conductor (`triage`) is the only judge. |
 | **Insight** | PyAutoInsight | Owns inference campaign intent and pending domain tasks, the cross-project inference instance registry, versioned `inference-summary` read contract, ingest receipts and evidence dashboard. Projects own execution, producers and raw samples; Cortex owns scientific run records, observations and human conclusions; Mind owns bounded implementation lifecycle and repository claims. Never infers scientific acceptance from execution, ranks incompatible runs or submits compute on refresh. |
+| **Broca** | PyAutoBroca | Owns assistant evidence — evaluation history, collection receipts and the assistant upkeep board (public, so committed evidence must be publishable: no raw transcripts or private data). Public assistants keep their definitions and runners and never depend on Broca; Brain interprets the evidence and routes accepted changes through Mind. Missing or incomparable results never establish a pass, and Broca never launches model or HPC campaigns. |
 | **DNA** | PyAutoDNA | Owns software-stack specifications, environment inventory receipts, compatibility decisions, upgrade campaigns and adoption history across local, RAL and CI. Brain coordinates updates; Heart owns validation/readiness; Hands owns release execution/provenance; Nerves owns runtime compatibility. Package repositories remain authoritative for requirements. Collection never upgrades an environment or treats missing observations as passing evidence. |
 | **Nerves** | PyAutoNerves | The configuration/serialization layer (`autonerves`) — layered config with overrides, the workspace↔library version handshake, `test_mode`, FITS/JSON I/O. Connects the organism's conventions to every library; the base layer the scientific libraries all import. |
 | **Gut** | PyAutoGut | Owns the lifecycle of *condemned self-material* — stale branches, stashes, dead code/tests. Holds each as a durable, recoverable git ref through a transit window and **voids** it on a sweep. The storage mirror of Memory (retention ↔ release); the hygiene conductor drives it, as vitals reads Heart. |
@@ -118,8 +119,8 @@ human types); let faculties multiply behind them.
 
 New capability grows as a **faculty** (cheap: one directory, one doc, one
 script), not as a repo. A new organ costs an `AGENTS.md`, install wiring, a body-map row and boundary prose — it must earn that by
-owning state or effects no existing organ can. Five capabilities have earned
-organ status that way. Configuration/signalling is the **Nerves**
+owning state or effects no existing organ can. Capabilities keep earning
+organ status that way; the first five were these. Configuration/signalling is the **Nerves**
 (PyAutoNerves), the base config/serialization layer every library imports — new
 config surfaces belong there, not in a new organ. Keeping track of what is
 true is the **Cortex** (PyAutoCortex), the second organ to earn it: it owns

agents/conductors/intake/_intake.py

@@ -41,7 +41,7 @@
     WORK_TYPES, HUMAN_REVIEW, MANUAL_ONLY_WORK_TYPES,
     LIBRARY_REPOS, WORKSPACE_REPOS, ORGANISM_REPOS, KNOWN_REPOS,
     RISK_KEYWORDS, AMBIGUITY_KEYWORDS, normalise_repo, norm_work_type,
-    declared_header, LANE_VALUES,
+    declared_header, LANE_VALUES, _body_map_specs,
     declared_inline, effective_difficulty, strip_declarations, _hits,
     effective_consequence, effective_unattended, effective_review_minutes,
     policy as _sizing_policy, BODY_MAP_PATH,
@@ -781,6 +781,12 @@ def _next_action(proposed, confidence, work_type=None):
 # prompt with no hits, or filed on a box with no PyAutoMemory, gets no line at
 # all: an empty `Memory:` would read as "nothing is known", which is a claim
 # this has no standing to make.
+#
+# Nor does a prompt that targets an organ (Brain, Mind, Heart, ...: the body
+# map's `category: organ` repos, i.e. ORGANISM_REPOS). Its subject is the
+# organism's own machinery, so there is no science question to answer, and the
+# digest only ever matched on shared vocabulary ("profile", "model") — every
+# organ prompt filed got galaxy-wiki citations that had to be hand-removed.
 MEMORY_CITATIONS = 3
 MEMORY_QUERY_WORDS = 200
 
@@ -811,9 +817,37 @@ def memory_citations(mind: Path, title: str, body_text: str,
     return [page["page"] for page in d["pages"]][:limit]
 
 
-def _with_memory(header: str, pages: list) -> str:
-    """`Memory: a.md; b.md` under the optional keys, or the header unchanged."""
-    if not pages:
+def _organs_dir_repos() -> frozenset:
+    """Normalised names of body-map repos checked out under `organs/`."""
+    try:
+        specs = _body_map_specs()
+    except (OSError, KeyError, ValueError):
+        return frozenset()
+    return frozenset(normalise_repo(name) for name, spec in specs.items()
+                     if str(spec.get("path", "")).startswith("organs/"))
+
+
+_ORGANS_DIR_REPOS = _organs_dir_repos()
+
+
+def _cites_memory(target: str | None) -> bool:
+    """Whether a prompt with this `Target:` key carries a `Memory:` line.
+
+    False for an organ target: see the block comment above. "Organ" here also
+    covers any repo whose body-map `path` sits under `organs/` (e.g. the
+    umbrella landing repo, `category: project` but organism machinery, not
+    science).
+    """
+    key = normalise_repo(target or "")
+    return key not in ORGANISM_REPOS and key not in _ORGANS_DIR_REPOS
+
+
+def _with_memory(header: str, pages: list, target: str | None = None) -> str:
+    """`Memory: a.md; b.md` under the optional keys, or the header unchanged.
+
+    Unchanged when there are no pages, or when `target` is an organ repo.
+    """
+    if not pages or not _cites_memory(target):
         return header
     line = "Memory: " + "; ".join(pages)
     lines = header.split("\n")
@@ -841,9 +875,11 @@ def write_prompt(mind: Path, decision: dict, body_text: str, source_note: str):
     # touches no filesystem, and a citation is only worth computing for a
     # prompt that is actually going to exist. The decision carries the header
     # that was written, so `--json` reports the file's real contents.
-    cites = memory_citations(mind, decision.get("title", ""), body_text)
+    target = decision.get("target")
+    cites = (memory_citations(mind, decision.get("title", ""), body_text)
+             if _cites_memory(target) else [])
     decision["memory"] = cites
-    decision["header"] = _with_memory(decision["header"], cites)
+    decision["header"] = _with_memory(decision["header"], cites, target)
     # The input's own header block (and its heading) is already rendered in
     # the decision's header — do not echo it into the body (PyAutoBrain#472).
     _heading, header_lines, body = _split_header(body_text)

bin/project_discovery.py

@@ -326,7 +326,7 @@ def main() -> int:
     parser.add_argument('mode', choices=('write', 'check', 'starter-write', 'starter-check'))
     parser.add_argument('repos', nargs='*')
     parser.add_argument('--root', type=Path, default=Path(os.environ.get('PYAUTO_ROOT', '.')))
-    args = parser.parse_args()
+    args = parser.parse_intermixed_args()  # install.sh puts --root before the repo names
     try:
         if args.mode.startswith('starter-'):
             if args.repos:

board/AGENTS.md

@@ -70,6 +70,13 @@ verbatim. This adapter supports retained sources; it does not re-enable
 retired overnight collection. Legacy explicit diagnostic helpers remain for
 compatibility, separate from board publication.
 
+The board's one remaining GitHub leg is upkeep: key
+`search/issues?q=org:<org>+is:issue+is:open&per_page=1`
+(`open_issues_endpoint()`), value GitHub's search response verbatim (the leg
+reads `total_count`). Versions, community and resume are retired and not
+collected, so `bin/version_drift.sh` and the community scan need no seam of
+their own; a retired leg returns only through this same endpoint-keyed map.
+
 `BOARD_GH` and `BOARD_PAGES_BASE` support hermetic fixtures. A failed Pages
 read may use an owning repo's local published artifact; it must not manufacture
 freshness. Public artifacts must never leak private local paths or credentials.

board/_board.py

@@ -112,7 +112,7 @@ def derive_org(homes):
 # ---------------------------------------------------------------------------
 # The injection seam (--github-data)
 #
-# Eleven of this board's legs read GitHub through `gh api`, and a Claude Code
+# The legs of this board that read GitHub do so through `gh api`, and a Claude Code
 # remote session — the surface the board is actually read from on a phone — has
 # no `gh` at all. Its GitHub access is the `mcp__github__*` tool surface, which
 # is an AGENT capability: this file is a subprocess and cannot reach it, however
@@ -475,10 +475,22 @@ def extract_heart_performance(board, board_url=""):
     }
 
 
+def open_issues_endpoint(org):
+    """The upkeep leg's `gh api` endpoint — also its `--github-data` key.
+
+    Upkeep is the board's only remaining GitHub leg (versions, community and
+    resume were retired with the agents/workflows simplification and are not
+    collected at all), so this is the one key a gh-less gatherer must supply.
+    The response is GitHub's search payload verbatim; the leg reads
+    `total_count`.
+    """
+    return f"search/issues?q=org:{org}+is:issue+is:open&per_page=1"
+
+
 def collect_open_issues(org, degraded):
     """Total open issues across the org — the /issue_cleanup pointer count
     (the audit itself stays that skill's confirmation-gated job)."""
-    res = gh_json([f"search/issues?q=org:{org}+is:issue+is:open&per_page=1"])
+    res = gh_json([open_issues_endpoint(org)])
     if res is None:
         degraded.append("upkeep: open-issue count unavailable")
         return None

docs/_extra/organs/build.html

@@ -0,0 +1,6 @@
+<!doctype html>
+<html lang="en"><head><meta charset="utf-8">
+<title>Moved: Hands</title>
+<link rel="canonical" href="hands.html">
+<meta http-equiv="refresh" content="0; url=hands.html">
+</head><body><p>This page moved to <a href="hands.html">Hands — PyAutoHands</a>.</p></body></html>

docs/board-orchestration.md

@@ -20,7 +20,7 @@ Supply the lowercase `organ` key for the approved action heading. The shared
 `prompt_heading(organ, heading_id=None)` helper also serves existing owner-specific
 controls without changing their prompts. It selects fixed text and emphasizes only
 the organ name with `strong`; it never accepts arbitrary heading HTML. Titles for
-legacy callers remain escaped plain text. All thirteen organ headings live in
+legacy callers remain escaped plain text. Every organ heading lives in
 `board/_theme.py`; reuse that mapping rather than copying sentences into renderers.
 Keep each heading on one line with responsive typography; do not hide overflow or
 truncate words. Omit explanatory subtitles and preserve work links/copy controls.

docs/concepts/organism.md

@@ -5,7 +5,9 @@ working with libraries and projects. Their boundaries are load-bearing. The cano
 definition lives in the Brain repo as
 [ORGANISM.md](https://github.com/PyAutoLabs/PyAutoBrain/blob/main/ORGANISM.md)
 — one page every organ links to instead of restating. This page summarises
-it and adds the framework/instance distinction an adopter needs.
+it and adds the framework/instance distinction an adopter needs. The list of
+organs grows over time; when this table and ORGANISM.md disagree, ORGANISM.md
+and the Mind's `repos.yaml` (every repo with `category: organ`) win.
 
 | Organ | Repo | Job |
 |-------|------|-----|
@@ -19,7 +21,9 @@ it and adds the framework/instance distinction an adopter needs.
 | **Hands** | PyAutoHands | *Does* — packaging, tagging, notebook generation, PyPI releases. A pure executor: never re-derives a gate decision. |
 | **Pulse** | PyAutoPulse | *Feels* — the cross-project profiling registry, the versioned `profiling-summary` read contract, ingest receipts and dashboard. Project repos produce and hold timings; Pulse validates the exchange and links to them. The Brain's profiling conductor handles judgment. |
 | **Insight** | PyAutoInsight | Coordinates inference campaign intent and domain tasks; owns the inference summary contract, registry, receipts and evidence board. Projects execute and retain samples; Cortex records scientific conclusions; Mind holds implementation lifecycle and claims. |
+| **DNA** | PyAutoDNA | *Remembers what it is made of* — software-stack specifications, environment inventory receipts, support rationale and upgrade/adoption history. Brain coordinates upgrades, Heart validates, Hands releases, Nerves enforces runtime compatibility; collection never upgrades an environment. |
 | **Nerves** | PyAutoNerves | *Connects* — the configuration/serialization layer (`autonerves`): layered config, the workspace↔library version handshake, `test_mode`, FITS/JSON I/O. The base layer every library imports. |
+| **Broca** | PyAutoBroca | *Watches the assistants* — assistant evaluation history, upkeep evidence, collection receipts and the upkeep board. Brain interprets; Mind tracks fixes; the public assistants stay independent of it. |
 | **Gut** | PyAutoGut | *Sheds* — the lifecycle of condemned self-material (stale branches, stashes, dead code/tests): holds each as a durable, recoverable git ref through a transit window, then **voids** it on a sweep. The storage mirror of Memory (retention ↔ release); the hygiene conductor drives it, as vitals reads Heart. |
 
 Everything else — the libraries being developed, their example workspaces,
@@ -38,7 +42,7 @@ is the container, and a workspace is one familiar kind of project.
 These roles do not replace the {doc}`repository categories <../satellites>` or
 change their workflow and release gates. Profiling, inference and visualization
 repos illustrate the project role; Eyes, Pulse and Insight provide cross-project organ
-views over the visualization and profiling projects. The project-to-organ
+views over the visualization, profiling and inference projects. The project-to-organ
 boundary is described in {doc}`../organs/eyes` and {doc}`../organs/pulse`.
 
 ## The call chain
@@ -61,11 +65,15 @@ The organs split on one line that matters for adoption:
   pipelines. Domain facts appear only in declared config surfaces (tables
   and policy files, not logic), and a drift check — the
   {ref}`tenant firewall <tenant-firewall>` — keeps it that way.
-- **Instance organs — Mind, Cortex, Memory, Eyes, Ears, Pulse, Insight, Gut.** Committed
-  state, ledgers, knowledge, visualization and profiling registries and shed
-  material. These are *inherently yours*: an adopter never forks the upstream
-  Mind, Cortex, Memory, Eyes, Ears, Pulse, Insight or Gut content, they create their own repos with the same documented
-  shape.
+- **Instance organs — every other organ except Nerves.** Mind, Cortex,
+  Memory, Eyes, Ears, Pulse, Insight, DNA, Broca and Gut hold committed
+  state: ledgers, knowledge, visualization, profiling and inference
+  registries, stack records, assistant evidence and shed material. These are
+  *inherently yours*: an adopter never forks their upstream content, but
+  creates their own repos with the same documented shape.
+- **Nerves sits outside the split.** It is an organ that also ships a library
+  package (`autonerves`) the scientific libraries import. You adopt it only if
+  your own libraries want that config layer; see {doc}`../organs/nerves`.
 
 One more principle worth knowing before you read anything else:
 **one canonical page per fact.** Organ boundaries live in ORGANISM.md;

docs/conf.py

@@ -26,6 +26,7 @@
 language = "en"
 
 html_static_path = ["_static"]
+html_extra_path = ["_extra"]  # e.g. organs/build.html -> hands.html after the rename
 html_css_files = ["pyauto.css"]
 
 html_theme_options = {

docs/index.md

@@ -2,9 +2,9 @@
 
 PyAutoScientist is a working, opinionated reference implementation of an
 **AI-agent development organism** for human-led, natural-language software
-development: a set of ten git repositories — Brain, Mind, Cortex, Memory,
-Eyes, Ears, Heart, Hands, Pulse, Insight, Nerves, Gut — through which you lead a multi-repo project in
-plain English. You
+development: a set of git repositories called **organs** (Brain, Mind,
+Cortex, Memory and the rest; {doc}`the organism <concepts/organism>` lists
+them all) through which you lead a multi-repo project in plain English. You
 describe what you want; AI agents plan, implement, test, gate and release
 it; you make every judgment call.
 
@@ -60,9 +60,12 @@ organs/memory
 organs/eyes
 organs/ears
 organs/heart
-organs/build
+organs/hands
 organs/pulse
 organs/insight
+organs/dna
+organs/nerves
+organs/broca
 organs/gut
 ```
 

docs/organs/broca.md

@@ -0,0 +1,39 @@
+# Broca — PyAutoBroca
+
+**What it owns:** the evidence on how the AI assistants are doing. Broca keeps
+assistant evaluation history, upkeep evidence, collection receipts and the
+assistant upkeep board.
+
+**Repo:** [PyAutoLabs/PyAutoBroca](https://github.com/PyAutoLabs/PyAutoBroca)
+
+## The defining function: watching the assistants speak
+
+Broca's area is the part of the brain tied to producing language. This organ
+watches the organism's public assistants and records how well they work.
+Benchmarks and maintenance inventory are kept as separate kinds of evidence,
+and a missing or incomparable result never counts as a pass.
+
+## The boundary that matters
+
+The public assistants stay independent. They keep their own definitions and
+reproducible runners, anyone can use them on their own, and they never depend
+on Broca.
+
+Broca records; it does not judge or act.
+
+- **Brain** interprets the evidence and routes accepted changes through Mind.
+- **Mind** tracks the fixes.
+- **Cortex** keeps science records and **Heart** keeps readiness, as everywhere
+  else.
+
+Broca is a public repository, so everything committed to it must be suitable
+for publication. Raw transcripts and private project data are excluded. It
+does not launch model or HPC campaigns on its own, and its Actions workflow
+publishes only the rendered saved snapshot, without collecting new evidence.
+
+## For an adopter
+
+Broca is an **instance organ**: its evidence is about *your* assistants. If you
+publish assistants of your own, create a Broca with the same shape. The full
+boundary is the
+[Broca section of ORGANISM.md](https://github.com/PyAutoLabs/PyAutoBrain/blob/main/ORGANISM.md#broca--assistant-evidence).