@assuredloop/orca-bot-kit 0.18.0 → 0.20.0

package/README.md

@@ -453,8 +453,21 @@ up. Each run renews its own schedule, which Claude Code would otherwise end
 after a week, and `--on --at` puts it back if it lapses. A `/clear` in its tab
 leaves it running until the session is next restarted; after that, `--on --at`
 puts it back.
-`obk groom --compact` compacts its conversation between runs. Grooming on Codex
-comes later: on a Codex Bot Father, add the session with `--harness claude`.
+`obk groom --compact` compacts its conversation between runs.
+
+The grooming session is always a Claude Code one, even on a Codex Bot Father (add
+it with `--harness claude` there): it is the clock, and where the reports arrive.
+To have the grooming itself done on Codex, turn it on with `--run-on codex`, and
+give the runs their model and effort:
+
+```sh
+obk groom --bots /path/to/my-bots --on --at 04:00 --run-on codex --model gpt-6-sol --effort medium
+```
+
+Each fire then makes a temporary Codex session of the grooming session's, named
+`groom-<date and time>`, which does the grooming, mails its report back, and is
+retired; the grooming session sends the report on to the management session. A
+run an earlier fire left behind is retired at the next one.
 
 ## When something is wrong
 
@@ -485,6 +498,10 @@ What it looks for:
   brings it back; when the book names no conversation for it, restart would
   refuse, and the finding says to close the tab in Orca and run
   `obk up … --session`. A tab the kit cannot read is never called down;
+- **a Claude session on an address the kit did not make**: such as the bare
+  `<bot>.<session>` every fleet's sessions had before 0.10.0, which a session of
+  another fleet can write to. Its next start through the kit gives it one of its
+  own, so the finding names the `obk restart … --session` that does it;
 - **leftovers no book owns**: an Orca project inside your bots folder with no
   bot in it, a tab in a bot's project that the book does not name, a start
   prompt written for a session that has gone, a clone of a skills source you no
@@ -598,7 +615,9 @@ real Orca and the real harnesses on your own machine. No CI runner can do that,
 so these are run by hand before a change that touches Orca or a harness is
 merged. A system test touches only what it creates and removes it afterwards,
 except the mailbox Runs Orca gives no way to delete, which the run names; it
-never closes a tab it did not open.
+never closes a tab it did not open. The trust its harnesses write into your
+`~/.codex/config.toml` and `~/.claude.json` for the run's own throwaway folders
+is taken out again after the run, and named; nothing else in them is changed.
 
 They run on the machine the command is typed on, so the command will not start
 them by itself. On its own it says what it is about to drive — whose machine,

package/SETUP.md

@@ -194,10 +194,10 @@ waiting for:
 | On screen | Send | Which is | Proven on |
 |---|---|---|---|
 | Claude Code's folder trust list | `\x1b[B\r` | down, then return: its selection starts on **No, exit** | Claude Code 2.1.283 |
-| Codex's directory trust, `1. Trust and continue` (older: `1. Yes, continue`) | `\r` | return: its selection starts on **Trust and continue** | Codex 0.158.0 |
-| Codex's `Hooks need review` | `\x1b[B\r` | down, then return, to **Trust all and continue**; if the selection already starts there, return alone | Codex 0.158.0 |
-| Codex's `/new`: `Where should the new conversation run?` | `\r` | return: its selection starts on **Current checkout** (bot home) | Codex 0.158.0 |
-| Codex's update offer, `1. Update now` | `\r` | return: its selection starts on **Update now**, which accepts it | Codex 0.156.1; seen on 0.157.1 (offering 0.158.0) with the selection on **Update now**, not pressed; 0.158.0 has had no newer release to offer |
+| Codex's directory trust, `1. Trust and continue` (older: `1. Yes, continue`) | `\r` | return: its selection starts on **Trust and continue** | Codex 0.160.0 |
+| Codex's `Hooks need review` | `\x1b[B\r` | down, then return, to **Trust all and continue**; if the selection already starts there, return alone | Codex 0.160.0 |
+| Codex's `/new`: `Where should the new conversation run?` | `\r` | return: its selection starts on **Current checkout** (bot home) | Codex 0.160.0 |
+| Codex's update offer, `1. Update now` | `\r` | return: its selection starts on **Update now**, which accepts it | Codex 0.156.1; seen on 0.157.1 (offering 0.158.0) with the selection on **Update now**, not pressed; 0.158.0 offered 0.160.0, taken on 2026-10-02 by Bot Father's restarts |
 | `[oh-my-zsh] Would you like to update?` | `n` | they update their own shell | oh-my-zsh bf77e35 |
 | Claude Code's `Teach auto mode about your environment?` | `\x1b` | Esc, which cancels it (**Not now**). On 2.1.283 it is a form, not a list: `How you use Claude here`, `Also scan shell history`, `Also scan your other repos`, `Continue`, and `Enter to continue · Esc to cancel`. **Never send a return**: Enter is **Continue**, which starts a scan of the project, recent Claude sessions and, by default, the machine's shell history | Not yet proven: Esc cancels in Claude Code 2.1.283's code, not yet seen live (#370); the form was seen live twice on 2.1.283, in #261's live test and #239's live run 1 (#416); 2.1.278 showed a numbered list, answered with 2. Not now |
 
@@ -217,7 +217,7 @@ empty book and nothing says why.
 Codex's trust question applies to the **repository root**, which for a bot means
 the whole bots folder rather than the one bot. That is what they are agreeing to.
 
-Codex's `/new` question (seen on 0.156.1, 0.157.1 and 0.158.0) is always answered with the current checkout, which is the bot
+Codex's `/new` question (seen on 0.156.1, 0.157.1, 0.158.0 and 0.160.0) is always answered with the current checkout, which is the bot
 home. The kit never makes a git worktree, so
 `2. New worktree` is never the answer.
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@assuredloop/orca-bot-kit",
-  "version": "0.18.0",
+  "version": "0.20.0",
   "license": "MIT",
   "repository": {
     "type": "git",
@@ -35,7 +35,8 @@
     "yaml": "2.9.0"
   },
   "devDependencies": {
-    "@stryker-mutator/core": "10.0.0"
+    "@stryker-mutator/core": "10.0.0",
+    "smol-toml": "1.9.0"
   },
   "overrides": {
     "typed-rest-client": {

package/skills/obk-arch/NOTICE.md

@@ -29,6 +29,9 @@ below.
   hatches in the types, the lock reflex where nothing was meant to be shared,
   callers needing the internal rules), with deviations treated as signal and a
   few edge cases not condemning a design.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, `architect`'s `runner-prompt` further to the above.
   Grounding the design in what the code actually touches before sketching, and
   what a sketch is (types and signatures with unimplemented bodies, a reader
@@ -37,17 +40,26 @@ below.
   and trusting types inside; one source of truth per invariant, derived rather
   than kept in step; and asking what happens if an operation runs twice or
   crashes halfway.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, `figure-it-out` further to the above. Capturing the
   baseline before the change so the check reads as the old value against the
   new; verdicts of verified, not verified or inconclusive with inconclusive not
   being a pass; inspecting the artefact rather than a report of it; and
   suspecting the observation method when something passes too easily.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, `blast-radius` further to the above. Looking where a
   symbol search stops: the library's own source and its pinned version, when
   things run, the shape of what an interface returns, a column, a wire format,
   another language reading the same bytes, a feature flag; giving each risk a
   real likelihood and a real cost; and listing separately what was checked and
   cleared, since a search that finds nothing is still a result.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, `create-verification-skill`. The whole of making a thing
   runnable and drivable, which the owner asked to live inside this skill rather
   than have one of its own: interviewing the repository rather than the user
@@ -69,6 +81,9 @@ below.
   which that skill's own users praised most because an agent stops searching
   the codebase for where things are. Kept here in that reduced form rather than
   as its own generated directory.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, `principle-sequence-verifiable-units`, `figure-it-out`,
   `blast-radius`, `principle-foundational-thinking`, `principle-model-the-domain`,
   `principle-exhaust-the-design-space` and the `poteto-mode` planning playbooks.
@@ -90,6 +105,9 @@ below.
   cannot happen, ran it, reproduced it) with "listing the callers is not the
   job", the one fact a change is safe because of, and anything short of
   running it said rather than written up as settled.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **mattpocock/skills**, `codebase-design` with `DEEPENING` and
   `DESIGN-IT-TWICE`, `domain-modeling` with its `ADR-FORMAT`, `prototype` and
   `to-tickets`. The three questions to put to an interface (fewer ways in,
@@ -116,6 +134,9 @@ below.
   the exception for a change that has to happen everywhere; and a prototype
   being throwaway code that answers a question, with the polish skipped on
   purpose.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **mattpocock/skills**, `triage` with its `AGENT-BRIEF`, and `to-tickets`
   further to the above. Handing work over behavioural rather than procedural,
   what the system should do and not how, the builder exploring fresh and
@@ -132,6 +153,9 @@ below.
   it was turned down before, asking the reporter specific questions rather
   than for more information, and noting when a piece needs a person's
   judgement rather than being delegated.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 Left behind on purpose: the multi-model arena and its runners, fixed model
 choices, the phase todolist and its ceremony, tracker and pull-request
@@ -173,11 +197,15 @@ Words were copied only from the two sources whose licences waive all rights:
   one or two pages, since nobody reads large documents; records numbered in
   sequence with numbers never reused; and a superseded record kept, because it
   is still worth knowing that it was the decision.
+  Read: 2026-09-24 (reconstructed: read by this date, when the entry was
+  written; the day it was read was not recorded)
 - **MADR 4.0.0** (adr.github.io/madr; MIT or CC0-1.0 at the reader's choice,
   taken here under CC0). The people who decided and the people consulted as
   fields of the record, "rejected" among the statuses, consequences written as
   good and bad, and a way to confirm the decision is kept (MADR's
   Confirmation, here "Checked by").
+  Revision: adr/madr@2475fe1973f66a12aaf58a91d8fa7b42c0f5ea3d (the 4.0.0 release
+  this entry names)
 
 Taken as ideas only, in this skill's own words, because their licences do not
 allow the text to be reused (all rights reserved unless noted):
@@ -193,23 +221,38 @@ allow the text to be reused (all rights reserved unless noted):
   decision; the confidence level; and the definition of done (evidence, at
   least two options compared on criteria, challenged and agreed, written and
   shared, a plan to check and revisit it).
+  Read: 2026-09-24 (reconstructed: read by this date, when the entry was
+  written; the day it was read was not recorded)
 - **Martin Fowler**, bliki "Architecture Decision Record". Writing the record
   as a way to bring disagreement into the open; most important first; the
   confidence and what should trigger a re-evaluation; an accepted record never
   reopened, only superseded.
+  Read: 2026-09-24 (reconstructed: read by this date, when the entry was
+  written; the day it was read was not recorded)
 - **AWS Prescriptive Guidance**, "Using architectural decision records…". A
   rejected record kept with its reason so the topic is not argued again; an
   accepted record not changed, a change being a new record noted in the
   record's change history.
+  Read: 2026-09-24 (reconstructed: read by this date, when the entry was
+  written; the day it was read was not recorded)
 - **Microsoft Azure Well-Architected Framework**, "Maintain an architecture
   decision record" (Microsoft Learn terms of use). Not editing accepted records
   but superseding them and linking the two; a record not being a design guide,
   standing on its own with detail linked; consequences never hidden.
+  Read: 2026-09-24 (reconstructed: read by this date, when the entry was
+  written; the day it was read was not recorded)
 - **Joel Parker Henderson**, `architecture-decision-record` (CC BY-NC-SA 4.0,
   which does not fit this use). Dating a fact that will change.
+  Revision:
+  joelparkerhenderson/architecture-decision-record@b1de91256a57110b655ff3572ce1fc41fbc2dac6
+  (reconstructed: the source's head when this entry was written, 2026-09-24; the
+  revision actually read was not recorded)
 - **adr-tools** (Nat Pryce; the tool is GPL-3.0, the content it adds to a
   project CC BY 4.0). Superseding changes only the old record's status, to a
   link to the new one.
+  Revision: npryce/adr-tools@b3279baf9be2207d1a4f4bbd608fd0b591c72aee
+  (reconstructed: the source's head when this entry was written, 2026-09-24; the
+  revision actually read was not recorded)
 - **The GDS Way**, "Documenting architecture decisions", and **GOV.UK**,
   "Architectural Decision Record Framework" (both Open Government Licence
   v3.0), and **Google Cloud Architecture Center**, "Architecture decision
@@ -218,6 +261,8 @@ allow the text to be reused (all rights reserved unless noted):
   names the decision rather than the problem, a superseded status with a link
   to the new record, records kept findable (GDS Way); and a change carrying the
   previous decision and why it changed (Google).
+  Read: 2026-09-24 (reconstructed: read by this date, when the entry was
+  written; the day it was read was not recorded)
 
 What the owner said, and how it shaped this part (issue #264, 2026-09-24): a
 record that partly replaces an older one replaces it with a new one, and he

package/skills/obk-bot-building/NOTICE.md

@@ -44,6 +44,11 @@ hooks means (PRD 6.7).
 
 ## Ideas taken from material with no licence
 
+- **The Grok Bot pages**, docs.x.ai/grok-bot and x.ai/bot/marketplace, as the
+  research pack's `grok-bot-lessons.md` gathers them.
+  Read: 2026-09-19 (as recorded in grok-bot-lessons.md, which fetched them that
+  day)
+
 The research pack's `grok-bot-lessons.md` gathers how one hosted bot product is
 run. Its material comes from a knowledge base and from that product's own
 marketplace and documentation pages. No licence is attached to those pages, and

package/skills/obk-bot-building/SKILL.md

@@ -288,9 +288,9 @@ otherwise `/Applications/Orca.app/Contents/Resources/bin/orca`. A bare
 | On screen | Send | Which is | Proven on |
 |---|---|---|---|
 | Claude Code's trust list | `\x1b[B\r` | down, return: off **No, exit** | Claude Code 2.1.283 |
-| Codex's `Trust this folder?` | `\r` | return: it starts on yes | Codex 0.158.0 |
-| Codex's `Hooks need review` | `\x1b[B\r` | **Trust all and continue** | Codex 0.158.0 |
-| Codex's `Where should the new conversation run?` | `\r` | **Current checkout** | Codex 0.158.0 |
+| Codex's `Trust this folder?` | `\r` | return: it starts on yes | Codex 0.160.0 |
+| Codex's `Hooks need review` | `\x1b[B\r` | **Trust all and continue** | Codex 0.160.0 |
+| Codex's `Where should the new conversation run?` | `\r` | **Current checkout** | Codex 0.160.0 |
 | Codex's update offer, `1. Update now` | `\r` | **Update now**, where it starts | Codex 0.156.1 |
 | `[oh-my-zsh] Would you like to update?` | `n` | no: the user updates it | oh-my-zsh bf77e35 |
 | Claude Code's `Teach auto mode…` | `\x1b` | Esc: Not now; no return | Not yet proven, below |
@@ -302,7 +302,8 @@ return: on Codex's update offer that took the highlighted **Update now**,
 whatever the digit was. Codex's trust said `1. Yes, continue` on older
 versions. Its hooks review sometimes starts on `2` already: then return alone.
 Its update offer was proven on 0.156.1; 0.157.1 showed it with the selection on
-Update now, not pressed, and 0.158.0 has had no newer release to offer.
+Update now, not pressed, and 0.158.0 offered 0.160.0, taken on 2026-10-02 by
+Bot Father's restarts.
 Claude Code's `Teach auto mode` on 2.1.283 is a form, not a list, ending
 `Enter to continue · Esc to cancel`: never send it a return, since
 Enter is Continue, which scans the project, recent sessions and the machine's
@@ -313,10 +314,17 @@ question matters most. The kit's hook is how the book learns which
 conversation the session is running, and until it is answered the
 conversation has not started. Codex's trust applies to the repository root,
 which is the whole bots folder. Codex's `/new` may ask where the new
-conversation runs (0.156.1, 0.157.1 and 0.158.0 all did); the answer is
+conversation runs (0.156.1, 0.157.1, 0.158.0 and 0.160.0 all did); the answer is
 always the current checkout, the bot home, and never `2. New worktree`:
 the kit never makes a git worktree.
 
+The hooks review of a Codex temporary session you made yourself is the one
+exception: answer it with `obk temp trust-hooks --bots <bots> --name
+<session>`, not with the keys. It reads the screen, chooses **Trust all and
+continue** only if the review is there, and checks that the review went. A
+permission rule can allow that one command, where it cannot allow keys sent
+into any tab.
+
 Where the kit says no session came up, the shell swallowed the launch line,
 usually while it was asking its own question. Answer the shell, then close
 that one tab (`<orca> terminal close --terminal <handle> --tab`) and run

package/skills/obk-debugging/NOTICE.md

@@ -20,6 +20,9 @@ with thanks:
   measuring before fixing a slow thing, the regression test at a correct seam
   and "no correct seam is itself the finding", redacting secrets from anything
   shown, and the list to clear before calling it done.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Waza**, `hunt`, through the verbatim passages in the research pack. The
   gate sentence and its form ("I believe the root cause is X because Y", named
   to a file, function, line or condition), the rule that the cause must explain
@@ -33,6 +36,8 @@ with thanks:
   re-read the path, and stopping after three failed hypotheses. The owner
   marked the gate sentence as worth borrowing in his own notes; the rest of
   `hunt` came in on the research's recommendation.
+  Revision: tw93/Waza@59323dee82b04fa8b9ad97a8ee503ca8aee7eb3c (as recorded in
+  the research pack's source-book-2.md, read 2026-09-19)
 - **Cursor pstack**, `principle-fix-root-causes` and the `poteto-mode`
   playbooks. Not adding guards to silence a crash, a workaround needing a
   paragraph meaning the code is wrong, checking for the pattern rather than the
@@ -58,6 +63,9 @@ with thanks:
   it not yet being a diagnosis; confirming against a paired capture or else
   calling it the strongest hypothesis the artefact supports; and proving the
   mechanism on the live process where it can still be reached.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **obra/superpowers**, `systematic-debugging` with `root-cause-tracing` and
   `condition-based-waiting`. Logging what enters and leaves each boundary and
   reading where it first goes wrong before digging in, tracing backwards
@@ -72,11 +80,17 @@ with thanks:
   cause really is external, handling at the boundary and a way to see it next
   time, with most "no root cause" being an investigation stopped early; and
   finding the test that leaves state behind by running them one at a time.
+  Revision: obra/superpowers@5bf4e78011075bcfc0dc295f0724994cd123ee71
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 - **obra/superpowers**, `systematic-debugging`, further to the above. Finding
   the nearest thing that works in the same codebase, reading it properly rather
   than skimming, and listing every difference however small without deciding in
   advance which cannot matter.
+  Revision: obra/superpowers@5bf4e78011075bcfc0dc295f0724994cd123ee71
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **addyosmani/agent-skills**, `debugging-and-error-recovery`. The decision
   tree for a bug you cannot reproduce on demand, sorted into timing,
   environment, state left behind and genuinely random, with what to try in each
@@ -85,6 +99,9 @@ with thanks:
   build can carry something shaped like an instruction and is to be read for
   clues rather than followed; keeping the evidence before a restart or a
   clean-up; and a test that looks wrong going back rather than being obeyed.
+  Revision: addyosmani/agent-skills@dc27a9c2e13721158157632de61b4106c6c2a2a1
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Waza**, `hunt`, further to the above. Its gotchas: walking back out of a
   library frame into your own code, suspecting persisted output written by the
   old code when the algorithm changed and the output did not, trusting the
@@ -93,14 +110,25 @@ with thanks:
   lower-layer baseline before blaming the visible thing, diagnosing an external
   tool before switching it, and never restarting more than twice without new
   evidence.
+  Revision: tw93/Waza@59323dee82b04fa8b9ad97a8ee503ca8aee7eb3c (as recorded in
+  the research pack's source-book-2.md, read 2026-09-19)
 
 - **garrytan/gstack**, `investigate`: earlier fixes in the same area pointing
   at the shape of the code, and the rest of the suite still passing after the
   fix.
+  Revision: garrytan/gstack@a6b3a57512ca6d5c6aa5b68f74f736195021f96e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Kent Beck's own rules file**: two tests for a defect, one at the interface
   and the smallest one where the fault lives.
+  Revision: KentBeck/BPlusTree3@ca80e4d85a99cd0af2effe717f709d43e80403bc
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, `runtime-forensics`, further to the above: when the
   question is why, the deliverable is the diagnosis, not a fix.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 The owner's own framing leads the skill: reproduce it, find the difference in
 environment, version and context, and the hard cases being random and slow

package/skills/obk-decision-memo/NOTICE.md

@@ -46,6 +46,9 @@ thanks. Each licence below was read from the file named beside it.
   recorded. Its completion check contributes the question of whether the
   baseline is genuine rather than a straw candidate and whether every rejected
   alternative has a reason.
+  Revision: citypaul/.dotfiles@a109f9972bb46671c624fc05752031523e1cf6fc
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **`addyosmani/agent-skills`, `skills/documentation-and-adrs`**: MIT,
   `addyosmani_agent-skills/LICENSE`, "Copyright (c) 2025 Addy Osmani". Writing
   a decision down when it would be expensive to reverse. Matching the
@@ -56,6 +59,9 @@ thanks. Each licence below was read from the file named beside it.
   giving the reason it was rejected. And the lifecycle: do not delete or edit
   an old decision when things change, write the next one and say which one it
   replaces, because the historical context is the part worth keeping.
+  Revision: addyosmani/agent-skills@dc27a9c2e13721158157632de61b4106c6c2a2a1
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **`addyosmani/agent-skills`, `skills/idea-refine`**: same licence. The
   machinery for alternatives that are actually different: the lenses for
   generating them, which this skill carries as asking what the opposite would
@@ -73,6 +79,9 @@ thanks. Each licence below was read from the file named beside it.
   those words. Its "not doing, and
   why" is the reason this skill treats the dropped options as a part of the
   memo rather than an appendix to it.
+  Revision: addyosmani/agent-skills@dc27a9c2e13721158157632de61b4106c6c2a2a1
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **`addyosmani/agent-skills`, `skills/doubt-driven-development`**: same
   licence. Checking your own conclusion: name the decision in two or three
   lines, because if you cannot get it that small you have a subject rather than
@@ -82,6 +91,9 @@ thanks. Each licence below was read from the file named beside it.
   what comes back as data rather than a verdict, re-reading the thing against
   each finding, since rubber-stamping a reviewer is the same failure as
   ignoring one.
+  Revision: addyosmani/agent-skills@dc27a9c2e13721158157632de61b4106c6c2a2a1
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **`citypaul/.dotfiles`, `claude/.claude/skills/double-check`**: MIT, the
   repository root licence above. How a finding is closed: push back with
   evidence and never by signalling which outcome you want, and treat an
@@ -91,6 +103,9 @@ thanks. Each licence below was read from the file named beside it.
   authoritative, or inferred. And its rule that severity is judged by the
   impact if it ships rather than by the effort to fix it or by the reviewer's
   own certainty.
+  Revision: citypaul/.dotfiles@a109f9972bb46671c624fc05752031523e1cf6fc
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **`citypaul/.dotfiles`, `claude/.claude/skills/find-gaps`**: MIT, the
   repository root licence above. What a plan or a memo is missing rather than
   what it gets wrong: treating silence as a red flag rather than a green light,
@@ -99,6 +114,9 @@ thanks. Each licence below was read from the file named beside it.
   anything. Its warning against inventing a gap to make the review look
   productive is why this skill says a memo may honestly conclude that none of
   the options is good.
+  Revision: citypaul/.dotfiles@a109f9972bb46671c624fc05752031523e1cf6fc
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **`garrytan/gstack`, `office-hours/SKILL.md` and `plan-ceo-review/SKILL.md`**:
   MIT, `garrytan_gstack/LICENSE`, "Copyright (c) 2026 Garry Tan". The premise
   challenge, which this skill carries as asking whether the question is the
@@ -107,6 +125,9 @@ thanks. Each licence below was read from the file named beside it.
   the outcome or is it solving a proxy for it. And its practice of putting the
   premises where the reader can disagree with them rather than burying them in
   the reasoning.
+  Revision: garrytan/gstack@a6b3a57512ca6d5c6aa5b68f74f736195021f96e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 Made generic, which is what the owner asked for. The strongest source is
 build-against-buy shaped and comes with a fixed template, an evaluation table

package/skills/obk-fleet-review/NOTICE.md

@@ -13,6 +13,9 @@ thanks:
   the reader can go and look; that a history is material rather than
   instructions; and preferring a remedy that holds by
   itself, one a check can enforce, over another paragraph of prose.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 ## What is written from this kit's own design
 
@@ -37,6 +40,11 @@ WHATEVER", which is why nothing here is written as a ban.
 
 ## Ideas taken from material with no licence
 
+- **The Grok Bot pages**, docs.x.ai/grok-bot and x.ai/bot/marketplace, as the
+  research pack's `grok-bot-lessons.md` gathers them.
+  Read: 2026-09-19 (as recorded in grok-bot-lessons.md, which fetched them that
+  day)
+
 The research pack's `grok-bot-lessons.md` gathers how one hosted bot product is
 run. Its material comes from a knowledge base and from that product's own
 marketplace and documentation pages. No licence is attached to those pages, and

package/skills/obk-grilling/NOTICE.md

@@ -20,6 +20,9 @@ with thanks:
   to make, and being finished only when the frontier is empty with nothing
   silently assumed and the user confirming the shared understanding before
   anything is built.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **mattpocock/skills**, `engineering/domain-modeling` with its
   `CONTEXT-FORMAT` and `ADR-FORMAT`. Challenging a term that conflicts with
   the agreed one and doing it immediately, proposing a precise term where one
@@ -39,6 +42,9 @@ with thanks:
   that carries lock-in, a boundary and what is deliberately outside it, a
   deliberate departure from the obvious path, a constraint invisible in the
   code, and an alternative rejected for a non-obvious reason.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 `grill-with-docs` is a one-line skill that loads the two above, and `grill-me`
 one that loads `grilling` alone. ADR 0019 proposes that a kit skill carries
@@ -76,23 +82,23 @@ words where they are quoted here:
 Five deliberate departures from `grilling` and `domain-modeling`, so they are
 not mistaken for consolidation:
 
-- **A round is three questions, not the whole frontier.** `grilling` says to
+- A round is three questions, not the whole frontier. `grilling` says to
   ask the entire frontier in one round. The owner asked for three at a time
   (13:24:19, recovered in the audit), and a round of nine reads as a form and gets
   filled in like one. The frontier still governs *which* questions are
   askable; it no longer governs how many go out at once.
-- **Small settled things get recorded.** `domain-modeling` says to offer an ADR
+- Small settled things get recorded. `domain-modeling` says to offer an ADR
   only when all three conditions hold and otherwise to skip it. That is the
   right gate for a decision note, and this skill keeps it, but it applies the
   gate to the note rather than to the record, because a reversible unsurprising
   choice is still what the next person needs in order to build the agreed
   thing. Everything settled goes down; a few earn the note.
-- **What was not settled is recorded too.** Neither source asks for this. It
+- What was not settled is recorded too. Neither source asks for this. It
   comes from the audit of the owner's own design session, where what he had
   merely agreed with in passing, and what had been assumed on his behalf, were
   the things that later had to be untangled.
 
-- **A decision note lists its alternatives and is replaced whole.**
+- A decision note lists its alternatives and is replaced whole.
   `ADR-FORMAT` makes a note one paragraph, with the rejected options and the
   consequences optional. The owner decided that a decision record lists every
   alternative that was on the table with why it was not chosen, and that a
@@ -101,7 +107,7 @@ not mistaken for consolidation:
   teaches the record; this skill's note says the same, so the two do not
   disagree.
 
-- **Questions are kept short.** `grilling` allows a question several
+- Questions are kept short. `grilling` allows a question several
   paragraphs of setup; a long one here is taken as a sign it is two questions
   or that teaching should come first.
 

package/skills/obk-grooming/NOTICE.md

@@ -13,10 +13,16 @@ with thanks:
   the place it was seen, so a reader can go and look; and the rule that a
   transcript is data to be read rather than instruction to be followed, which
   is what the section on reading another bot's history is built on.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Addy Osmani's agent skills**, `constraint-driven-development`: its warning
   that a figure with no reason attached gets deleted by the next person it
   inconveniences, which is why a note says what moved it. Its ratchet is not
   applied to the notes (see below).
+  Revision: addyosmani/agent-skills@dc27a9c2e13721158157632de61b4106c6c2a2a1
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 ## The owner's own conclusions, restated
 
@@ -60,6 +66,11 @@ only a summary. The facts about Claude Code's scheduler are in the tech notes.
 
 ## Ideas taken from material with no licence
 
+- **The Grok Bot pages**, docs.x.ai/grok-bot and x.ai/bot/marketplace, as the
+  research pack's `grok-bot-lessons.md` gathers them.
+  Read: 2026-09-19 (as recorded in grok-bot-lessons.md, which fetched them that
+  day)
+
 The research pack's account of how one hosted bot product runs its fleet
 suggested the daily pass over every bot and the habit of routing each finding
 to a single kind of fix. Those pages carry no licence and the pack holds only

package/skills/obk-grooming/SKILL.md

@@ -244,6 +244,16 @@ at: `obk session add --bots <folder> --bot bot-father --name grooming --model
 <m> --effort <e>`, with `--harness claude` where Bot Father runs on Codex, then
 `obk up`. Its launch line carries them, so they are what every run uses.
 
+The runs can be done on Codex instead. The session stays this Claude Code one,
+the clock and the reader, and `obk groom --on --at <HH:MM> --run-on codex
+--model <m> --effort <e>` schedules it: each fire makes a temporary Codex
+session, `groom-<date and time>`, with that model and effort, which does the
+pass with this skill, mails its report to the grooming session and is retired
+by it; the grooming session sends the report on to the management session. A
+run an earlier fire left behind is retired at the next fire. If you are such a
+run, do the pass, send the report as your task says, and stop: schedule
+nothing and make no session.
+
 Do not schedule a pass nobody has watched. `obk groom --now` runs it once in its
 tab; read what it produced, and fix the run rather than the report. Then, after
 one explicit yes from the user who has read a run, `obk groom --on --at <HH:MM>`

package/skills/obk-handoff/NOTICE.md

@@ -15,6 +15,9 @@ with thanks:
   From its `docs/productivity/handoff.md`: the next agent takes the note as a
   contract and will not re-check it, so read it before handing it over and
   downgrade anything only assumed.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor pstack**, the `poteto-mode` `pause-safely` playbook: stopping at a
   safe boundary by finishing or backing out of the current step and never
   stopping mid-edit in a known-broken state, starting nothing new and stopping
@@ -26,6 +29,9 @@ with thanks:
   conversation, the commits and whether the tree is clean, and the first action
   on resume ("this is a pause, not a final report"). Also its rule that a pause
   is explicit: on "keep going", do not pause.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 Two additions are not from either source, both from an acceptance run of this
 skill: commit your own work and only your own, since the durable-work

package/skills/obk-personal-facilitation/NOTICE.md

@@ -22,6 +22,8 @@ contribute a piece each.
   scheduling for a reminder and verifying what was set, a line in a to-do file
   or a saved skill not being a scheduled reminder, and reporting a missing host
   capability instead of claiming something will run later.
+  Revision: guwenqing/orca-bot-kit@51a42a3e61711b0f33b00482657499de9afee16e (the
+  commit the `git show` above names)
 - **The research pack's `grok-bot-lessons.md`**, our own read of the owner's
   knowledge base on running a personal helper bot. "36 drafts queued, 0 sent":
   finish every reversible step and stop at the irreversible one (BNK-2920 §09,
@@ -29,13 +31,20 @@ contribute a piece each.
   (BNK-2921); here scoped to a routine that does work, since a plain reminder
   has nothing to run by hand. The time zone and the end date in a routine's
   checklist.
+  Revision: none (the kit's own research pack, not a published source)
 - **`mattpocock/skills`, `skills/in-progress/loop-me`** (MIT): do the most you
   can before involving the person, so they are asked once, late; and hand back
   something decision-ready that points at the thing rather than reproducing it.
+  Revision: mattpocock/skills@c55ee46073ed923f86ce59a5eb3b6d895095d1b7
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 - **Cursor `plugins`, `third_party/x/skills/x-chat`** (MIT, `third_party/x/
   LICENSE`, "Copyright (c) 2026 Cursor"): outbound text needs the person's
   approval unless they already said to send it ("reply that I'll be there"),
   and a vague "check my inbox" is not that.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 Written differently from the old skill, on purpose: no fixed path (PRD 7.1
 assumes no way of working); a vague item stays on the list and is asked about,

package/skills/obk-recall/NOTICE.md

@@ -37,6 +37,9 @@ thanks:
   to re-derive it, and a "verify from scratch" pass means treating it as
   untrustworthy when it is not; and a passing prior self-report is not the
   proof.
+  Revision: cursor/plugins@6ed0f7a9504f577d7529064103cecce9be7dfc5e
+  (reconstructed: the research pack's 2026-09-20 clone the skills were built
+  from; the revision actually read was not recorded)
 
 Made generic, which is what the owner asked for. The original names a fixed
 transcript path, a particular set of hosted sources and a parallel fan-out on