1 files changed, 54 insertions, 0 deletions
diff --git a/archive/task-archive.org b/archive/task-archive.org
index 9d2bfc6..e52563d 100644
--- a/archive/task-archive.org
+++ b/archive/task-archive.org
@@ -1664,3 +1664,57 @@ Touches four synced template workflows and needs a curation pass on the best-pra
 Drafted [[file:docs/design/2026-06-16-encourage-kb-contribution-spec.org][the KB-contribution spec]]: four light workflow prompts (startup nudge, triage-intake + inbox-zero end-of-flow reminders, an early wrap-up reflection feeding the existing KB receipt) plus one Craig-authored best-practices node curated from Ahrens / Matuschak / org-roam guidance. Five open sub-decisions filed as decisions-as-TODO in the spec.
 *** 2026-06-20 Sat @ 23:29:10 -0400 Spec ratified + built
 Craig ratified all five decisions (2026-06-20) and added D6 — a read-side startup consult-nudge surfacing project-relevant KB node titles, the counterpart the original write-only design lacked. Built all of it: the best-practices node (=~/org/roam/agents/20260620232112-agent-kb-best-practices.org=), startup's two Phase C nudges (consult + contribute, gated on the roam clone), the conditional capture reminders in triage-intake + inbox-zero, and the early wrap-up reflection feeding the existing receipt. Commits 76e5559 (workflows + spec) and the related lint checker f6dde4e. Trigger for the build: receipt data showed "promoted 0 / consulted no" across recent sessions.
+** DONE [#C] Bash/shell language bundle :feature:
+CLOSED: [2026-06-23 Tue]
+:PROPERTIES:
+:CREATED: [2026-06-23 Tue]
+:END:
+Built =languages/bash/= the same session it was filed: bash.md + bash-testing.md rules, a shellcheck PostToolUse validate hook (covers =.sh=, =.bash=, and extensionless shell scripts by shebang; 8 bats tests), a shellcheck pre-commit githook, settings.json wiring, gitignore-add.txt, and a "Bash/shell project" CLAUDE.md. shfmt left out of the blocking path on purpose (shell has no canonical style). Makefile test target now discovers =languages/*/tests/*.bats=.
+
+No =languages/= bundle fits a shell-heavy project. archangel (437 =.sh= files) and archsetup are bash projects with nothing that matches; installing elisp/python gives them the wrong language rules. Build a =languages/bash/= bundle on the elisp/go pattern: =claude/rules/bash.md= (style — =set -euo pipefail=, quoting, =[[ ]]=, trap/cleanup) + =bash-testing.md= (bats conventions), a PostToolUse validate hook (=shellcheck= on edited =.sh=), a =githooks/pre-commit= running shellcheck on staged shell files, =settings.json= wiring, =gitignore-add.txt=, and its own =CLAUDE.md= headed "Bash/shell project." Urgency dropped 2026-06-23: install-lang now seeds the language-neutral default CLAUDE.md when a bundle ships none, so a bash project no longer gets a mislabeled "Elisp project" header — the bundle is now the accurate-rules win, not a mislabel fix. From archangel 2026-06-23 ([[file:docs/design/2026-06-23-install-lang-claude-md-gap.org][handoff]]).
+** DONE [#B] Consolidate inbox/triage workflows + scheduled inbox check :chore:
+CLOSED: [2026-06-23 Tue]
+:PROPERTIES:
+:CREATED: [2026-06-23 Tue]
+:END:
+Built per the Ready spec: =process-inbox= + =monitor-inbox= + =inbox-zero= merged into one =inbox.org= engine (shared core + process/monitor/roam modes + the interactive =auto inbox zero= =/loop= mode); =triage-intake= and =no-approvals= stay separate. Callers repointed (INDEX, protocols, startup Phase C, wrap-up Step 3), old files deleted, stale-ref grep clean, workflow-integrity + sync-check + full suite green. The fully-unattended =/schedule= cron pass is vNext (see the =[#D]= task above). [[file:docs/specs/inbox-workflow-consolidation-spec.org][spec]].
+** DONE [#C] inbox-zero: delete empty roam entries on triage :feature:
+CLOSED: [2026-06-23 Tue]
+:PROPERTIES:
+:CREATED: [2026-06-23 Tue]
+:END:
+Done in commit 3da2725 (empty-entry sweep folded into Phase D's reconcile, after capture-guard + pull, with the claimed-item removal) and carried into the consolidated =inbox.org= roam mode (Phase B =empty= bucket + Phase D sweep). From the roam inbox 2026-06-23.
+** DONE [#C] Surface cross-project dependencies first in what's-next :feature:spec:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:CREATED: [2026-06-24 Wed]
+:END:
+Tasks that depend on another project can sit for ages when the dependency is low-priority or needs its own spec process — e.g. wrap-teardown depends on =.emacs.d= for the =ai-term= companion. Craig's proposal (roam 2026-06-24): (1) an org-tag marking a task as blocked-by / depends-on another project (pick a short tag name); (2) several ways to bind dependencies into the what's-next (=open-tasks.org=) decision tree so blocked-by-dependency tasks surface first; (3) review the what's-next workflow as a whole, since many projects use it.
+
+Built 2026-06-24 (tag name =:blocked:=, Craig's pick): the =:blocked:= tag + =:BLOCKED_BY: <project>: <what>= property convention in =todo-format.md=, and =open-tasks.org= Next Mode now excludes =:blocked:= tasks from the cascade and surfaces them in a dedicated "Blocked on other projects" section with an =inbox-send= nudge offer. Applied live to the wrap-teardown task above. Commits feat(tasks) cross-project-dependency.
+** DONE [#C] Task-audit: consolidate adjacent / related tasks :feature:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:CREATED: [2026-06-24 Wed]
+:END:
+The task-audit workflow should also consider combining related tasks when they're adjacent, so a spread-out effort reads as one whole. Craig's example (roam 2026-06-24): the agent-agnostic / agent-source work could collapse into one item, or at least a parent task with the related ones as children.
+
+Built 2026-06-24: =task-audit.org= Phase C.5 reads the open-task set, spots semantic clusters by judgment, and proposes per cluster either a merge (same-work members fold into one) or a parent-with-children grouping (related-but-distinct), applied only on Craig's confirm — broader than Phase C's exact-duplicate fold. Commit feat(task-audit) consolidate.
+** DONE [#B] Anki deck name from #+TITLE :bug:quick:solo:
+CLOSED: [2026-06-24 Wed]
+:PROPERTIES:
+:CREATED: [2026-06-22 Mon]
+:LAST_REVIEWED: 2026-06-24
+:END:
+flashcard-to-anki.py's =default_deck_name= returns =input_path.stem= (the filename), so every deck built through =flashcard-sync= (which passes no =--deck=) is named after the file, not the curated =#+TITLE=. =flashcard-review.org= already documents the intended behavior ("the #+TITLE line drives the Anki deck name"); the script never matched it. Fix: =default_deck_name(input_path, org_text)= scans for a =#+TITLE:= line (case-insensitive, trimmed) and returns it, basename fallback when absent; =main()= passes the already-read =org_text=. Edited script + test ready (validated, 29 pass); the staging =.py= files were removed after the fix landed (see below), rationale kept: [[file:docs/design/2026-06-21-anki-titlefix-proposal.org][proposal]]. Apply to both =.ai/scripts/= and =claude-templates/.ai/scripts/=, sync-check + make test. Migration caveat: deck ID derives from the name, so decks previously built without =--deck= land as new decks on next import (old basename-named decks keep history, delete by hand). Coordinate with "Reconcile flashcard multi-tag tooling into canonical" below — both edit =flashcard-to-anki.py=, build together to avoid conflicting edits. Shared-asset, review-gated. From home 2026-06-21.
+
+Done 2026-06-24 (commit 060a938): applied the pre-staged script + test red-to-green (5 new =#+TITLE= tests, 29 pass total), synced both script dirs, full suite green. The two redundant staging =.py= files removed, the rationale proposal kept.
+** CANCELLED [#C] Morning ops orchestrator pilot — read-only :feature:
+CLOSED: [2026-06-24 Wed 05:46]
+:PROPERTIES:
+:CREATED: [2026-06-11 Thu]
+:LAST_REVIEWED: 2026-06-15
+:END:
+A scheduled headless morning run chaining the existing pieces: startup checks, the triage-intake scan, a system health check — producing the prep doc plus a report and a notify ping, with all remediation propose-only. Staged adoption from the 2026-06-11 insights report's "Self-Healing Daily Ops Orchestrator": read-only first; promote individual routine remediations to auto only after each has a track record. Known blockers to design around: headless MCP auth (interactively-authenticated servers are absent in cron runs) and the consent boundary (triage Phase D, anything destructive).
+
+The triage limb can reuse triage-intake's *auto mode* (added 2026-06-15, see [[file:.ai/workflows/triage-intake.org]]) — its accumulate-don't-mutate sweep is the propose-only behavior this orchestrator wants. Auto mode itself runs in-session (inherited MCP auth); the orchestrator is the durable headless schedule, so the headless-auth blocker above is the part still on this task to solve.