refactor: read/write Mind completions via complete/ records

AGENTS.md

@@ -182,7 +182,7 @@ humans invoke identically, so behaviour isn't re-derived from prose each time.
 - **`agents/faculties/memory/`** — *recalls what the organism knows*: given a
   topic, returns a **cited digest** (ranked pages + snippets) over
   PyAutoMemory's sub-wikis, `autolens_assistant`'s skills/wiki, and Mind's
-  `complete.md` history. Grep-ranked at query time — no indexes, no layout
+  `complete/` records. Grep-ranked at query time — no indexes, no layout
   coupling, never a page dump; an empty digest is the honest answer. Privacy
   seam: PyAutoMemory citations never reach public user-facing output.
 - **`agents/faculties/samplers/`** — *the motor faculty*: expertise in how the

ORGANISM.md

@@ -16,7 +16,7 @@ organs plan, build, test and release it, and you make every judgment call.
 | **Brain** | PyAutoBrain | Figures out *how* — reasoning, planning, decomposition, routing; hosts the specialist agents. Owns **no state, no health checks, no execution mechanics**. |
 | **Heart** | PyAutoHeart | Determines whether the organism is healthy. `pyauto-heart readiness` is the **authoritative** GREEN/YELLOW/RED "is it safe to release?" gate. An observer: never writes into other repos, never triggers Build. |
 | **Hands** | PyAutoBuild | 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. |
-| **Memory** | PyAutoMemory | Long-term knowledge — *what the science says* (literature wikis, concepts, bibliographies). Operational history — *what the organism did* — lives in Mind (`complete.md`, issues), not here. |
+| **Memory** | PyAutoMemory | Long-term knowledge — *what the science says* (literature wikis, concepts, bibliographies). Operational history — *what the organism did* — lives in Mind (the `complete/` records, issues), not here. |
 | **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. |
 
 The scientific libraries (PyAutoConf, PyAutoFit, PyAutoArray, PyAutoGalaxy,

agents/conductors/intake/AGENTS.md

@@ -85,7 +85,7 @@ schema — light structure over free-form prose.
 | **census** | `intake census` | inventory every filed prompt (work-type/target/difficulty/status + hygiene flags); always read-only |
 | **dashboard** | `intake dashboard` | render the census as the Mind **backlog** page; `--apply` writes `PyAutoMind/dashboard.md` |
 | **formalise** | `intake formalise [prefix]` | retroactively header the prompts census flags — derive the missing fields, insert in place, prose untouched; `--apply` writes |
-| **reconcile** | `intake reconcile [prefix]` | rank backlog prompts that look already-shipped (vs `complete.md` / `issued/`); always read-only — retiring stays human |
+| **reconcile** | `intake reconcile [prefix]` | rank backlog prompts that look already-shipped (vs the `complete/` records / `active/`); always read-only — retiring stays human |
 
 Census/dashboard are the Mind *backlog* view — deliberately distinct from
 Heart's `/health status` health view (see "must never do"). The prompt-taxonomy
@@ -101,7 +101,7 @@ moves or deletes a file.
 Reconcile exists because a prompt's `Status:` header is **not** a completeness
 signal — formalise preserves an existing Status verbatim, so shipped work can
 still read `Status: planned`. It cross-references each backlog prompt against
-`complete.md` (path references + `## header` topic overlap), `issued/`
+the `complete/` records (path references + `## header` topic overlap), `active/`
 basenames, and hand-set Status values, then ranks suspects (high/medium/low)
 with the evidence shown. The final verification — the target repo's git log /
 merged PRs — and the retirement itself stay human.

agents/conductors/intake/INTAKE_TAXONOMY.md

@@ -125,7 +125,7 @@ only under `--apply` and re-runs are no-ops (nothing left missing).
 status went stale (a header `Status:` is display metadata — formalise preserves
 an existing value verbatim, so `Status: planned` can outlive the work by
 months). Per backlog prompt it collects four Mind-local signals — a
-`complete.md` line referencing its path (follow-up/restore/parked wording
+completion-record line referencing its path (follow-up/restore/parked wording
 downgrades it to likely-open), a duplicate basename in `issued/`, token overlap
 with a completed task's `## header`, and a hand-set Status — then ranks
 suspects high/medium/low with the evidence shown. Always read-only: the final

agents/conductors/intake/_intake.py

@@ -592,8 +592,8 @@ def emit_formalise(res: dict):
     "the a an of to in for and or is are be with on by via from into as at it "
     "this that use using make add new fix update support get set can we i you "
     "our my need should will when once each all its".split())
-# Wording in a complete.md reference line that marks the prompt as a deferred
-# follow-up (still open) rather than the shipped task itself.
+# Wording in a completion-record reference line that marks the prompt as a
+# deferred follow-up (still open) rather than the shipped task itself.
 _FOLLOWUP_WORDS = ("follow", "restore", "parked", "remain", "blocked", "later",
                    "next step", "next-step", "deferred")
 
@@ -606,23 +606,28 @@ def _tokens(s: str) -> set:
 def reconcile(mind: Path, prefix: str = "") -> dict:
     """Rank backlog prompts that look already-shipped, for a human to retire.
 
-    Mind-local signals per prompt: a legacy complete.md line referencing its
+    Mind-local signals per prompt: a completion-record line referencing its
     path (follow-up wording downgrades it), a duplicate basename in `active/`
     or the `complete/` archive, token overlap with a completed task's header /
     archive record, and a hand-set Status the formalise pass deliberately
     preserved. Never writes anything.
     """
     c = census(mind)
-    comp = mind / "complete.md"
-    comp_lines = (comp.read_text(encoding="utf-8", errors="replace").splitlines()
-                  if comp.is_file() else [])
-    headers = [(ln[3:].strip(), _tokens(ln[3:].replace("-", " ")))
-               for ln in comp_lines if ln.startswith("## ")]
     comp_dir = mind / "complete"
     comp_files = ([p for p in comp_dir.rglob("*.md")
-                   if "archive" not in p.parts and p.name != "AGENTS.md"]
+                   if "archive" not in p.parts
+                   and p.name not in ("AGENTS.md", "index.md")]
                   if comp_dir.is_dir() else [])
     comp_names = {p.name for p in comp_files}
+    # reference lines + `## <slug>` topic headers now live inside the dated
+    # records (the monolithic complete.md ledger was retired — issue #81)
+    comp_lines: list = []
+    for p in comp_files:
+        comp_lines.extend(
+            p.read_text(encoding="utf-8", errors="replace").splitlines())
+    headers = [(ln[3:].strip(), _tokens(ln[3:].replace("-", " ")))
+               for ln in comp_lines
+               if ln.startswith("## ") and ln[3:].strip() != "Original prompt"]
     headers += [(f"complete/{p.relative_to(comp_dir)}",
                  _tokens(p.stem.replace("_", " "))) for p in comp_files]
     active = mind / "active"
@@ -664,7 +669,7 @@ def reconcile(mind: Path, prefix: str = "") -> dict:
         if best[0] >= 0.40 or len(best[2]) >= 3:
             score = best[0]
             findings.append(("topic-overlap",
-                             f"complete.md '## {best[1]}' "
+                             f"completion record '{best[1]}' "
                              f"(shared: {', '.join(sorted(best[2]))})"))
 
         if r["status"] not in ("-", "formalised"):
@@ -697,8 +702,8 @@ def emit_reconcile(res: dict):
     print(f"== Intake reconcile: {len(res['suspects'])} suspect(s) of "
           f"{res['scanned']} scanned ==")
     if not res["suspects"]:
-        print("  backlog reconciles clean against the complete/ archive, "
-              "complete.md and active/.")
+        print("  backlog reconciles clean against the complete/ records "
+              "and active/.")
     for s in res["suspects"]:
         print(f"[{s['confidence']:>6}] {s['path']}")
         for f in s["findings"]:

agents/conductors/intake/intake.sh

@@ -25,7 +25,7 @@
 #                                census flags (word-vomit is intent, not defect);
 #                                --apply writes the headers in place
 #   intake reconcile [prefix]    rank backlog prompts that look already-shipped
-#                                (vs complete.md / issued/); always read-only —
+#                                (vs the complete/ records / active/); always read-only —
 #                                retiring a prompt stays human
 #
 # Flags (place before the subcommand; both default OFF):

agents/faculties/memory/AGENTS.md

@@ -4,7 +4,7 @@
 > *consult*, not a front door you drive. It *recalls what the organism knows*:
 > given a topic or question, it returns a **cited digest** — pointers into
 > PyAutoMemory's sub-wikis, `autolens_assistant`'s skills/wiki pages, and
-> Mind's operational history (`complete.md`) with matching snippets — and
+> Mind's operational history (the `complete/` records) with matching snippets — and
 > stops. It never dispatches, never mutates, never bulk-loads a wiki into
 > context.
 
@@ -23,16 +23,16 @@ synthesises. The digest is pointers + evidence, never wholesale content:
 memory.sh "delaunay regularization prior work"
   -> ranked pages (path · hit count · 1-2 matching snippet lines each)
      across: PyAutoMemory/<sub-wiki>/ · autolens_assistant/{skills,wiki}/ ·
-     PyAutoMind/complete.md sections
+     PyAutoMind/complete/ records
 ```
 
 - **No indexes, no embeddings, no new infra** — grep + the wikis' own
   structure, the same as a careful human session, kept deterministic.
 - **No layout coupling** — sub-wikis are discovered at query time
   (`*_wiki/` directories); when Memory grows a wiki, the faculty sees it with
   no edits (the standing "do not couple to Memory's internal layout" rule).
-- Operational history stays Mind's: `complete.md` is *read* here as a recall
-  surface, but the boundary prose in ORGANISM.md is unchanged.
+- Operational history stays Mind's: the `complete/` records are *read* here as
+  a recall surface, but the boundary prose in ORGANISM.md is unchanged.
 
 ## Privacy seam (hard rule)
 

agents/faculties/memory/_memory.py

@@ -3,7 +3,7 @@
 
 The **memory faculty** is a read-only opinion sink: given a topic/question, it
 greps the organism's knowledge surfaces — PyAutoMemory's sub-wikis,
-autolens_assistant's skills/wiki pages, and Mind's complete.md history — and
+autolens_assistant's skills/wiki pages, and Mind's complete/ records — and
 returns a **cited digest**: ranked pages with matching snippets. The consulting
 agent reads only the listed pages and synthesises; this script never dumps
 content, never writes, and builds no index.
@@ -41,8 +41,16 @@ def surfaces(memory: Path | None, assistant: Path | None, mind: Path | None):
             if root.is_dir():
                 for f in root.rglob("*.md"):
                     yield f"autolens_assistant/{sub}", assistant.parent, f
-    if mind and (mind / "complete.md").is_file():
-        yield "PyAutoMind/complete.md", mind.parent, mind / "complete.md"
+    comp = mind / "complete" if mind else None
+    if comp and comp.is_dir():
+        # the dated records are Mind's completion history (complete.md retired,
+        # issue #81); archive/ holds non-record material, index/AGENTS are meta
+        for f in comp.rglob("*.md"):
+            if f.name in ("index.md", "AGENTS.md"):
+                continue
+            if "archive" in f.relative_to(comp).parts:
+                continue
+            yield "PyAutoMind/complete", mind.parent, f
 
 
 def score_file(f: Path, terms: list[str]):

agents/faculties/memory/memory.sh

@@ -3,7 +3,7 @@
 # read-only reasoning capability). It recalls what the organism knows.
 #
 # Greps the knowledge surfaces — PyAutoMemory sub-wikis, autolens_assistant
-# skills/wiki, PyAutoMind complete.md — and emits a cited digest (ranked pages
+# skills/wiki, PyAutoMind complete/ records — and emits a cited digest (ranked pages
 # + matching snippets). The consulting agent reads only the listed pages and
 # synthesises. Read-only: never writes, never dumps whole pages, never invents
 # context when the surfaces are empty. Privacy seam: PyAutoMemory citations

docs/concepts/workflow.md

@@ -23,10 +23,10 @@ idea  →  prompt file  →  issue  →  worktree  →  PR(s)  →  merge  →
 4. **`ship_*`** gates the finish: test suites, the review faculty, and the
    Heart readiness verdict. GREEN ships; YELLOW needs an explicit human
    acknowledgement; RED stops. One PR per repo, always.
-5. **Merge is always human.** After merge, the task entry moves from
-   `active.md` to `complete.md` — the organism's operational history.
+5. **Merge is always human.** After merge, the task leaves `active.md` and a
+   dated `complete/` record is written — the organism's operational history.
 
-The registry files (`active.md`, `planned.md`, `complete.md`) are shared
+The registry files (`active.md`, `planned.md`, the `complete/` records) are shared
 state in the Mind repo, so any machine or session — laptop, web, CI — can
 read the current picture and resume an in-flight task with no handoff
 ceremony.

docs/example.md

@@ -38,7 +38,7 @@ Development touched PyAutoArray and PyAutoGalaxy; the ship gate ran both
 test suites, the review faculty, and Heart's verdict; one PR landed per
 repo; the workspaces then gained a demonstration script in a follow-up
 workspace task. The registry recorded every state transition, and the task
-retired to `complete.md`.
+retired to its dated `complete/` record.
 
 That is the whole pitch: the input was a paragraph of intent, and every
 step from there — planning, isolation, testing, gating, releasing,

docs/organs/memory.md

@@ -13,7 +13,7 @@ Memory is **pull-only, on demand**. No agent loads it automatically; the
 Brain's memory faculty consults it when a task is scientific, returning a
 cited digest (pages + snippets) rather than bulk content. Operational
 history — what the organism *did* — deliberately lives in the Mind
-(`complete.md`, issues), not here: Memory is what the domain says, Mind is
+(the `complete/` records, issues), not here: Memory is what the domain says, Mind is
 what the organism did.
 
 ## The pieces

docs/organs/mind.md

@@ -2,7 +2,7 @@
 
 **What it owns:** intent and workflow state. Every task starts here as a
 plain-English markdown prompt; the registry files track what is in flight
-(`active.md`), queued (`planned.md`) and done (`complete.md`); and
+(`active.md`), queued (`planned.md`) and done (the dated `complete/` records); and
 `repos.yaml` — the **body map** — is the single source of repo identity for
 the whole organism.
 
@@ -18,7 +18,7 @@ the whole organism.
   repo or domain. Prompts are free-form prose; an optional light header
   (`Type:` / `Difficulty:` / `Autonomy:` / `Priority:`) gives both humans
   and the Brain a glance-level summary.
-- **The registry** (`active.md` / `planned.md` / `complete.md`) is the
+- **The registry** (`active.md` / `planned.md` / the `complete/` records) is the
   shared task state — worktree claims, issue links, status, resume notes.
   It is what makes the workflow machine-independent.
 - **The body map** (`repos.yaml`) declares every repo's GitHub home,

skills/WORKFLOW.md

@@ -11,7 +11,7 @@ the individual skill files can stay short.
 The organs and boundaries are defined once in
 [`../ORGANISM.md`](../ORGANISM.md). What the workflow skills need to know on
 top of that: Mind owns the workflow **state** (`active.md` / `planned.md` /
-`complete.md`, the prompt taxonomy), and Build owns **no dev-workflow skills**
+the `complete/` records, the prompt taxonomy), and Build owns **no dev-workflow skills**
 — it is the release/packaging executor only.
 
 A workflow skill reasons through **Brain**, gates ship through **Heart**, records
@@ -97,7 +97,7 @@ execution tier — no manual model toggling.
 (`start_library`/`start_workspace`), release triage (`review_release`);
 identifying affected repos, drafting the commit message and full PR body
 (`## API Changes` / `## Scripts Changed`), workspace-impact analysis, the
-library-first merge gate, the merge decision, `active.md` / `complete.md`
+library-first merge gate, the merge decision, `active.md` / completion-record
 updates, and final issue comments. In `pre_build`: validating clean `main`,
 asking for the minor version, printing the summary.
 
@@ -202,13 +202,14 @@ shared task state, so any environment reads it and continues an in-flight task.
 Workflow skills read and write Mind state via **workspace-root-relative** paths
 that resolve from any sibling repo:
 
-- `PyAutoMind/active.md`, `PyAutoMind/planned.md`, `PyAutoMind/complete.md` (ledgers)
-- Prompt **files** advance `draft/ → active/ → complete/<YYYY>/<MM>/` in lockstep
-  with the ledgers; `PyAutoMind/scripts/lifecycle.py` owns the moves (`move`) and
-  drift-checks them (`check`). See `PyAutoMind/complete/AGENTS.md` (issue #71).
+- `PyAutoMind/active.md`, `PyAutoMind/planned.md` (live ledgers)
+- Prompt **files** advance `draft/ → active/ → complete/<YYYY>/<MM>/`; the dated
+  record IS the completion ledger (`complete.md` retired 2026-07-16, issue #81).
+  `PyAutoMind/scripts/lifecycle.py` owns the writes (`record`, `move`) and
+  drift-checks them (`check`). See `PyAutoMind/complete/AGENTS.md` (issues #71/#81).
 - `source PyAutoMind/scripts/prompt_sync.sh` → `prompt_sync_push "<msg>"` (commit+push registry)
 
-`active.md` task schema, `complete.md` schema, and the prompt taxonomy
+`active.md` task schema, the completion-record schema, and the prompt taxonomy
 (`draft/feature/<target>/`, `draft/bug/<target>/`, …) are documented in
 `PyAutoMind/README.md`. **Under `draft/`, the first folder is the work type, the
 second is the target repo/domain.**

skills/health/health.md

@@ -11,7 +11,7 @@ Shared routing context: `PyAutoBrain/skills/COMMANDS.md`.
 |------------|--------------|---------------|
 | `/health` | The Brain **Health conductor** loop with you: assess (vitals → Heart) → triage → dispatch a validation leg → re-judge, until GREEN. | `bin/pyauto-brain health` |
 | `/health check` | One-shot **green-light sweep**: sync each repo's `main`, run library pytest + workspace smoke, report a pass/fail matrix. | `PyAutoHeart/skills/health_sweep/reference.md` |
-| `/health status` | **Active-work dashboard**: what's in flight across the repos (reads `active.md` / `planned.md` / `complete.md`), conflicts, idle repos. | `PyAutoHeart/skills/pyauto-status/reference.md` |
+| `/health status` | **Active-work dashboard**: what's in flight across the repos (reads `active.md` / `planned.md` / the `complete/` records), conflicts, idle repos. | `PyAutoHeart/skills/pyauto-status/reference.md` |
 | `/health full` | **Release-run dashboard**: what the last PyAutoBuild release-prep run produced (per-workspace pass/fail/timing, slowest scripts, failure tracebacks). | `PyAutoHeart/skills/pyauto-status-full/reference.md` |
 | `/health worktrees` | **Worktree/task dashboard**: every task worktree under `~/Code/PyAutoLabs-wt/`, cross-referenced with `active.md` — which parallel tasks are in flight and which repos each holds. | `PyAutoHeart/skills/worktree_status/reference.md` |
 

skills/intake/intake.md

@@ -40,7 +40,7 @@ dev workflow (issue, branch, plan). Do not bypass the Brain.
   (then regenerate the dashboard).
 - `bin/pyauto-brain intake reconcile [prefix]` — ranks backlog (`draft/`)
   prompts that look already-shipped (cross-referenced against the `complete/`
-  archive — or the legacy `complete.md` until retired — and `active/`; a stale
+  records — the sole completion ledger since `complete.md` retired, #81 — and `active/`; a stale
   hand-set `Status:` is a signal, never proof). Always read-only: verify each
   suspect against the target repo's git log / merged PRs, then retire it to the
   `complete/` archive by hand (it is already done).