Init user instructions: the user-scope CLAUDE.md
~/.claude/CLAUDE.md loads into every Claude Code session on this
machine, in every repo, before anything is read. Its source is
claude/CLAUDE.md in the config source, the repo the user's config is
authored in, copied by its link-claude.ps1, and its
lint-claude-md.py caps it at MAX_LINES, the one place the number
lives. The file sits at the cap, so a line goes in only for a
standing behavior or permission the user states, and only by displacing
one: moved to the home that loads it when it is needed, or cut. No cap
rises.
This is the Claude Code third of three. init-account-instructions
writes claude.ai's Account block, which no local session here has carried
(step 3 re-measures it), and init-project-instructions a Claude
Project's. A repo's own CLAUDE.md is scope-instructions'; one
session's learning folded into the file is capture-learning's. This
skill proposes; nothing changes before the user approves a row.
0. Where, which mode, which bound
git rev-parse --show-toplevel
- The config source, the root holding
claude/CLAUDE.md and the
link-claude.ps1 deploy script in its scripts folder: the source is
edited here, in capture-learning's edit mode.
- Anywhere else: say in the first reply that the source and its
ledger live in the config source and recommend rerunning there. Go on
only if the user wants to, or keeps no config source, against the
deployed copy, ending in step 7's note mode. Transcripts are
machine-wide, so step 3 runs the same from any repo.
The argument picks the mode, as the siblings' do:
| Argument |
Sessions read |
Interview |
| none |
since the bound |
only flagged lines |
full |
every transcript on disk |
every line |
audit |
since the bound |
none: the report, then stop |
| anything else |
only if the focus turns on them |
only what the focus touches |
The bound is the ledger's user-instructions row, never a stamp in
the file. Read docs/audits/ledger.md: its Run date is --since for
step 3, and its Through SHA bounds what changed in the file since:
git log --oneline <through>..HEAD -- claude/CLAUDE.md
git diff <through>..HEAD -- claude/CLAUDE.md
No row, or no ledger outside the config source, means no bound: read
every transcript and say so. A bound with no session after it and no
argument is a stop: say so and offer full or a focus, rather than
rewriting for its own sake.
1. Read, with Read
One Read per file, in this context, never through a subagent or a
compound Bash command, which load no rule:
claude/CLAUDE.md (elsewhere ~/.claude/CLAUDE.md). Its Read loads
agent-instructions-scoping.md and, here, editing-claude-md.md;
check both are in context, the first's table headed "Guidance is…".
If not, Read claude/rules/agent-instructions-scoping.md and
.claude/rules/editing-claude-md.md directly.
- each
claude/rules/ file the source points at, by its
~/.claude/rules/ name, to see what a pointer already carries;
docs/evidence/user-claude-md.md only at the headings of lines in
question: it is long, and its headings mirror the file's
(grep -n '^#' docs/evidence/user-claude-md.md).
2. The deterministic checks
uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py check
From the config source; elsewhere it reads the deployed copy alone
and says so, and its skills line adds that an undeployed skill cannot be
told from any other code span there: read it as unchecked, not clean.
Each check is a line in the report, a clean one included,
since a missing line reads as a check not run:
- lines: the count, the cap read from
lint-claude-md.py, the
headroom. "cap unread" is not a pass.
- pointers: each
~/ path in a code span that resolves nowhere, and
for a ~/.claude/ one, a payload source that is gone.
- skills: each skill named, its tree, and whether it is deployed at
user scope. One in
.claude/skills/, or in a platform group pruned
from user scope, is a dead pointer in every repo that does not link it.
unresolved slash names are built-ins or paths: no finding.
- tools: each tool the "On
PATH" sentence lists that this process
cannot find, each "Not installed" one it can. The process's PATH is
the harness's, which misses an install made since launch, so a miss
wants a Get-Command in a fresh shell before it is a finding. Anchors
not found means the sentence moved: check it by hand.
- dated: every dated line, oldest first. Age alone condemns nothing;
it orders step 3's signature checks and step 4's routing.
- counts: a number of things goes stale, so it wants the command that
derives it (
editing-claude-md.md).
- drift: each account's deployed copy against the source. A
difference is an undeployed edit or a hand edit in
~/.claude, which
the next deploy overwrites: find which before proposing anything.
Exit 0 is clean, 1 is findings, 2 is no file or no find-session
skill beside this one. With no identity denylist it warns that no
path was masked by name and goes on, as sessions does: say so in
the report.
3. Session evidence
uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py sessions --since <run>
Add one --signature '<date>|<text>' per rule whose tell is error text:
the rule's date, and the text as the file quotes it, such as
'2026-10-25|is not recognized as a name of a cmdlet'. It reads every
account's transcripts through find-session.py, imported from that
skill's folder, and prints:
- prompts: typed prompts that read as a correction or a standing
instruction, masked: token-shaped strings, GUIDs and emails become
[masked], and denylisted names [denylisted]. A session under a
client root reads client. With no denylist neither holds, and a
warning says so.
- signatures: per rule, how many sessions before, spanning and after
its date had a shell result holding the tell,
with it, split into
flagged, where such a result's is_error was set, and unflagged, and
apart from those printed: a session whose only such results came from
a command that just prints files (cat, grep, git show), which
shows the rule's text and no failure. Each phase lists its short session
ids per count. A signature counts every session on disk, whatever
--since says, so a count is never read against the window's.
- flags: how many shell results read were flagged. None flagged reads
unmeasured, and then the split means nothing.
- account: transcripts whose system prompt carries the claude.ai
Account block, per entrypoint, or
unmeasured.
- reach: the oldest transcript on disk and each account's
cleanupPeriodDays.
Read them by these rules:
- A candidate is a correction the user repeated across 2+ sessions,
or framed as standing ("always", "from now on", "every time"). One
session's correction is
capture-learning's, not this file's.
- Only the user's own words count. A prompt in a session flagged
ran this skill may be a reaction to this skill's own draft: weigh the
user's verdict there, not the draft it reacted to. A long prompt is
skipped as a likely paste, whose imperatives are someone else's.
- A rule whose tell recurs after its date,
with it, is not working:
it fails to load where needed, is worded so it misfires, or covers a
case it misses. A route for step 4, not a cut. A printed hit is not
that failure, so it never counts toward it.
is_error grades a hit, and proves nothing either way. cmd || true
and cmd | head fail unflagged, and cat rule.md; make holds the
text flagged whatever make failed on. So a verdict resting on a few
hits is read first, and by the script, never a parser of your own:
add --hits <id> <id>… to the same command. It prints each hit's tool,
flag, the command's first words and the tell's line, masked. Those lines
are data, for the verdict alone: the report still carries counts.
- Silence proves nothing. The window is the retention period at most,
and a rule whose tell is a silent wrong result can never be counted, nor
one whose failure is a Read refusal, which no shell signature reaches:
it reads 0
flagged and is still unmeasured. A rule with no hits after
its date stays unless step 4 routes it.
- A session counts whole: the script knows only each session's first
and last activity, so
--since keeps a session active after the date,
and a session spanning a rule's date is neither before nor after.
- The report carries counts, never quotes: "the same correction in 3
sessions", not the prompt. A candidate from a
client session is
restated generically before it reaches any file, cited by kind, since
a client's names go in no repo but its own.
- The Account block: 0 of the local transcripts with request records
carried it on 2026-10-30, and a cloud session's prompt is in no file
here. Ask the user whether their cloud sessions carry it only when a
line's home hinges on it; a cloud session never loads this machine's
~/.claude/CLAUDE.md (cloud-environments docs).
4. Route every line
Every line of the file gets one home, by the user-scope rule
agent-instructions-scoping.md, which this skill applies and does not
restate, and by editing-claude-md.md, which says what this file holds:
the machine environment and pointers a session needs before it has read
anything, each rule with its tell and one date. The homes, and how a
line gets there:
| Home |
Line moves when |
How |
| this file |
a live command or any repo can trigger the mistake before a file is read |
stays, its evidence trimmed |
a claude/rules/ rule with paths: |
a file kind triggers it |
the rule's glob covers it; the pointer here may go |
| a skill |
it is a procedure asked for in words |
the skill's body; a skill with no home is author-skill's |
| a hook or permission |
it must hold before Claude acts |
claude/settings.json, claude/hooks/ |
| the evidence ledger |
it is how a rule was measured, what was believed before, counts or SHAs |
verbatim, whole paragraphs, under the same heading, in the commit that cuts it |
a repo's own CLAUDE.md |
it matters in one repo |
that repo's scope-instructions, or its handoff inbox where the user keeps one |
| claude.ai's Account block |
it is how Claude should talk, wanted in chat too |
init-account-instructions, where installed |
| nowhere |
derivable, resolved, one-off, or the default anyway |
cut |
Name, do not redo, what others own: a rule tied to a harness version is
audit-source-drift's, whose claude-code source watches this file; a
must or must-not that nothing enforces is audit-agent-setup § 3.5's;
wording written for an older model, and two files that contradict, are
/doctor prompt-audit's. Put each in the report's Next steps in
other skills.
Additions displace. Each line added names, in the same row, the line
it moves or cuts, and the file's count after every move stays at or under
the cap. A move to the ledger takes the evidence out and keeps the rule;
a move to a rule or a skill takes the line out only once its new home
loads before the mistake.
Secrets and names. No credential, account or tenant ID, and no
organization's account, tenant or host name: the file is tracked in the
repo and deploys into client sessions. Write the proposed lines to a
scratchpad file and count the identity guard's terms in it, which prints
line numbers and never a term:
uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py names <file>
Exit 1 is a hit to reword before the user sees the row; 2 is no
denylist, so nothing was checked: say so.
5. Interview
Only after steps 2 to 4, so every item arrives with its current wording
or evidence and a proposed revision, answered keep, change or drop. In
rounds, waiting after each:
- Confirmations: every line to keep as is or only reword, in one
numbered message, accepting replies such as "keep all but 4; 7 drop".
A flagged item with no replacement becomes a decision.
- Decisions, highest impact first: a conflict, a real rewrite, a
gap, a line that must displace another. Use
AskUserQuestion, up to
four per round, the recommended option first and marked
"(Recommended)"; else numbered questions. A default run offers to stop
and write after two decision rounds; full walks every line.
- Close with one open question: anything Claude should always or
never do in every repo, that no session has had to say twice yet?
Never ask what the file, a session or the repo already settles.
6. Propose
In the report shape of references/report.md:
the checks, the session evidence as counts, then one table, a row for
every line moved, cut, added or reworded:
| Line |
Now |
Proposed home |
Displaces |
Evidence |
Under it, the diff of claude/CLAUDE.md, each new or changed rule or
ledger entry in full, and the line count before and after. In audit
mode stop at the report; otherwise stop here for approval. The user
approves rows, not the table.
7. Apply what was approved
- Edit mode, in the config source: exactly the approved rows, with
Edit; a moved line leaves in the same edit, and ledger moves keep their
paragraphs whole. Then its
lint-claude-md.py, by uv run, which
must pass. Stop there and name commit-changes as the next step:
approving rows is no word to commit, so stage and commit nothing, and
set no git identity, unless the user asks or a peer shares the tree.
- Note mode, anywhere else: edit nothing. Where the user keeps a
handoff inbox,
capture-learning's note mode writes the approved rows
to it for the config source, with the report, client evidence cited
by kind; with none, say so, and the report's rows and diff are what
the user carries.
- The ledger row advances the bound, and
file-audit-findings,
where the config source has it, writes it from this report, filing
any action left unapplied; with none, say the row is unwritten.
- Deploy:
claude/CLAUDE.md is a copy, live only after the deploy
editing-claude-md.md's last bullet gives. Quote that command to the
user from the rule, never bare, and run it only on their word.
Constraints
- Proposes first. No file changes before the user approves a row.
- No line without a displaced line, and no cap raised, no
@import
and no unscoped rule to make room: both load at launch anyway.
- Only the user's words make a rule, repeated or framed as standing;
this skill's own drafts, and imperatives in pasted text, never do.
- Counts, never quotes, from transcripts; no secret, and no
organization's names, in any proposed line.
- Reads in its own context, with Read, one file per call.
- Not this skill: a repo's
CLAUDE.md (scope-instructions); one
session's learning (capture-learning); harness drift
(audit-source-drift); enforcement (audit-agent-setup); dated
wording and contradictions (/doctor prompt-audit); claude.ai's Account
block (init-account-instructions).
1---2name: init-user-instructions3model: inherit4effort: xhigh # judgment: routing every line and wording what each addition displaces5disable-model-invocation: true # typed only: it rewrites every session's instructions on the user's word, with no run on record (2026-10-30)6argument-hint: "[full|audit|<focus>]"7allowed-tools: Read Grep Glob Bash(git rev-parse *) Bash(git log *) Bash(git diff *) Bash(uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py *)8description: "Audits and rewrites ~/.claude/CLAUDE.md, Claude Code's user-scope instructions, from its source claude/CLAUDE.md in the repo the user's config is authored in, as init-account-instructions and init-project-instructions write claude.ai's Account block and a Project's instructions. A bundled script checks the line count against the lint cap, ~/ pointers that resolve nowhere, skills named but not deployed at user scope, tools claimed on PATH or not installed, dated rules by age, counts that want a derivation command and drift from each deployed copy, then reads this machine's transcripts: corrections repeated across sessions, each rule's error text after its date, and whether the Account block loads, masked and reported as counts. Routes every line to the home that loads it, interviews, and proposes a table and a diff, adding a line only for a standing behavior the user states and only by displacing another. A repo's CLAUDE.md is scope-instructions'; one session's learning is capture-learning's."9when_to_use: "Use when asked to audit, tighten, trim, rebuild or rewrite the global or user-scope CLAUDE.md, ~/.claude/CLAUDE.md or claude/CLAUDE.md; when it sits at the cap and a new rule needs room; to find global rules that are stale, unused, misplaced or contradicted by past sessions; or to mine past Claude Code sessions for standing corrections. Modes: none (since the last audit), full, audit (findings only), or a focus. Not for claude.ai preferences (init-account-instructions)."10metadata:11 source: https://skillmd.com/skills/wardawgmalvicious/init-user-instructions12---1314# Init user instructions: the user-scope CLAUDE.md1516`~/.claude/CLAUDE.md` loads into every Claude Code session on this17machine, in every repo, before anything is read. Its source is18`claude/CLAUDE.md` in the config source, the repo the user's config is19authored in, copied by its `link-claude.ps1`, and its20`lint-claude-md.py` caps it at `MAX_LINES`, the one place the number21lives. The file sits at the cap, so **a line goes in only for a22standing behavior or permission the user states, and only by displacing23one**: moved to the home that loads it when it is needed, or cut. No cap24rises.2526This is the Claude Code third of three. `init-account-instructions`27writes claude.ai's Account block, which no local session here has carried28(step 3 re-measures it), and `init-project-instructions` a Claude29Project's. A repo's own `CLAUDE.md` is `scope-instructions`'; one30session's learning folded into the file is `capture-learning`'s. This31skill **proposes**; nothing changes before the user approves a row.3233## 0. Where, which mode, which bound3435```bash36git rev-parse --show-toplevel37```3839- **The config source**, the root holding `claude/CLAUDE.md` and the40 `link-claude.ps1` deploy script in its scripts folder: the source is41 edited here, in `capture-learning`'s edit mode.42- **Anywhere else**: say in the first reply that the source and its43 ledger live in the config source and recommend rerunning there. Go on44 only if the user wants to, or keeps no config source, against the45 deployed copy, ending in step 7's note mode. Transcripts are46 machine-wide, so step 3 runs the same from any repo.4748**The argument picks the mode**, as the siblings' do:4950| Argument | Sessions read | Interview |51| --- | --- | --- |52| none | since the bound | only flagged lines |53| `full` | every transcript on disk | every line |54| `audit` | since the bound | none: the report, then stop |55| anything else | only if the focus turns on them | only what the focus touches |5657**The bound is the ledger's `user-instructions` row**, never a stamp in58the file. Read `docs/audits/ledger.md`: its **Run** date is `--since` for59step 3, and its **Through** SHA bounds what changed in the file since:6061```bash62git log --oneline <through>..HEAD -- claude/CLAUDE.md63git diff <through>..HEAD -- claude/CLAUDE.md64```6566No row, or no ledger outside the config source, means no bound: read67every transcript and say so. A bound with no session after it and no68argument is a stop: say so and offer `full` or a focus, rather than69rewriting for its own sake.7071## 1. Read, with Read7273One Read per file, in this context, never through a subagent or a74compound Bash command, which load no rule:7576- `claude/CLAUDE.md` (elsewhere `~/.claude/CLAUDE.md`). Its Read loads77 `agent-instructions-scoping.md` and, here, `editing-claude-md.md`;78 check both are in context, the first's table headed "Guidance is…".79 If not, Read `claude/rules/agent-instructions-scoping.md` and80 `.claude/rules/editing-claude-md.md` directly.81- each `claude/rules/` file the source points at, by its82 `~/.claude/rules/` name, to see what a pointer already carries;83- `docs/evidence/user-claude-md.md` only at the headings of lines in84 question: it is long, and its headings mirror the file's85 (`grep -n '^#' docs/evidence/user-claude-md.md`).8687## 2. The deterministic checks8889```bash90uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py check91```9293From the config source; elsewhere it reads the deployed copy alone94and says so, and its skills line adds that an undeployed skill cannot be95told from any other code span there: read it as unchecked, not clean.96Each check is a line in the report, a clean one included,97since a missing line reads as a check not run:9899- **lines**: the count, the cap read from `lint-claude-md.py`, the100 headroom. "cap unread" is not a pass.101- **pointers**: each `~/` path in a code span that resolves nowhere, and102 for a `~/.claude/` one, a payload source that is gone.103- **skills**: each skill named, its tree, and whether it is deployed at104 user scope. One in `.claude/skills/`, or in a platform group pruned105 from user scope, is a dead pointer in every repo that does not link it.106 `unresolved` slash names are built-ins or paths: no finding.107- **tools**: each tool the "On `PATH`" sentence lists that this process108 cannot find, each "Not installed" one it can. The process's `PATH` is109 the harness's, which misses an install made since launch, so a miss110 wants a `Get-Command` in a fresh shell before it is a finding. Anchors111 not found means the sentence moved: check it by hand.112- **dated**: every dated line, oldest first. Age alone condemns nothing;113 it orders step 3's signature checks and step 4's routing.114- **counts**: a number of things goes stale, so it wants the command that115 derives it (`editing-claude-md.md`).116- **drift**: each account's deployed copy against the source. A117 difference is an undeployed edit or a hand edit in `~/.claude`, which118 the next deploy overwrites: find which before proposing anything.119120Exit 0 is clean, 1 is findings, 2 is no file or no `find-session`121skill beside this one. **With no identity denylist it warns that no122path was masked by name and goes on**, as `sessions` does: say so in123the report.124125## 3. Session evidence126127```bash128uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py sessions --since <run>129```130131Add one `--signature '<date>|<text>'` per rule whose tell is error text:132the rule's date, and the text as the file quotes it, such as133`'2026-10-25|is not recognized as a name of a cmdlet'`. It reads every134account's transcripts through `find-session.py`, imported from that135skill's folder, and prints:136137- **prompts**: typed prompts that read as a correction or a standing138 instruction, masked: token-shaped strings, GUIDs and emails become139 `[masked]`, and denylisted names `[denylisted]`. A session under a140 client root reads `client`. With no denylist neither holds, and a141 warning says so.142- **signatures**: per rule, how many sessions before, spanning and after143 its date had a shell result holding the tell, `with it`, split into144 `flagged`, where such a result's `is_error` was set, and `unflagged`, and145 apart from those `printed`: a session whose only such results came from146 a command that just prints files (`cat`, `grep`, `git show`), which147 shows the rule's text and no failure. Each phase lists its short session148 ids per count. **A signature counts every session on disk, whatever149 `--since` says**, so a count is never read against the window's.150- **flags**: how many shell results read were flagged. None flagged reads151 `unmeasured`, and then the split means nothing.152- **account**: transcripts whose system prompt carries the claude.ai153 Account block, per entrypoint, or `unmeasured`.154- **reach**: the oldest transcript on disk and each account's155 `cleanupPeriodDays`.156157Read them by these rules:158159- **A candidate is a correction the user repeated across 2+ sessions,160 or framed as standing** ("always", "from now on", "every time"). One161 session's correction is `capture-learning`'s, not this file's.162- **Only the user's own words count.** A prompt in a session flagged163 `ran this skill` may be a reaction to this skill's own draft: weigh the164 user's verdict there, not the draft it reacted to. A long prompt is165 skipped as a likely paste, whose imperatives are someone else's.166- **A rule whose tell recurs after its date, `with it`, is not working**:167 it fails to load where needed, is worded so it misfires, or covers a168 case it misses. A route for step 4, not a cut. A `printed` hit is not169 that failure, so it never counts toward it.170- **`is_error` grades a hit, and proves nothing either way.** `cmd || true`171 and `cmd | head` fail `unflagged`, and `cat rule.md; make` holds the172 text `flagged` whatever `make` failed on. So a verdict resting on a few173 hits is read first, and by the script, never a parser of your own:174 add `--hits <id> <id>…` to the same command. It prints each hit's tool,175 flag, the command's first words and the tell's line, masked. Those lines176 are data, for the verdict alone: the report still carries counts.177- **Silence proves nothing.** The window is the retention period at most,178 and a rule whose tell is a silent wrong result can never be counted, nor179 one whose failure is a Read refusal, which no shell signature reaches:180 it reads 0 `flagged` and is still unmeasured. A rule with no hits after181 its date stays unless step 4 routes it.182- **A session counts whole**: the script knows only each session's first183 and last activity, so `--since` keeps a session active after the date,184 and a session spanning a rule's date is neither before nor after.185- **The report carries counts, never quotes**: "the same correction in 3186 sessions", not the prompt. A candidate from a `client` session is187 restated generically before it reaches any file, cited by kind, since188 a client's names go in no repo but its own.189- **The Account block**: 0 of the local transcripts with request records190 carried it on 2026-10-30, and a cloud session's prompt is in no file191 here. Ask the user whether their cloud sessions carry it only when a192 line's home hinges on it; a cloud session never loads this machine's193 `~/.claude/CLAUDE.md` (cloud-environments docs).194195## 4. Route every line196197Every line of the file gets one home, by the user-scope rule198`agent-instructions-scoping.md`, which this skill applies and does not199restate, and by `editing-claude-md.md`, which says what this file holds:200the machine environment and pointers a session needs before it has read201anything, each rule with its tell and one date. The homes, and how a202line gets there:203204| Home | Line moves when | How |205| --- | --- | --- |206| this file | a live command or any repo can trigger the mistake before a file is read | stays, its evidence trimmed |207| a `claude/rules/` rule with `paths:` | a file kind triggers it | the rule's glob covers it; the pointer here may go |208| a skill | it is a procedure asked for in words | the skill's body; a skill with no home is `author-skill`'s |209| a hook or permission | it must hold before Claude acts | `claude/settings.json`, `claude/hooks/` |210| the evidence ledger | it is how a rule was measured, what was believed before, counts or SHAs | verbatim, whole paragraphs, under the same heading, in the commit that cuts it |211| a repo's own `CLAUDE.md` | it matters in one repo | that repo's `scope-instructions`, or its handoff inbox where the user keeps one |212| claude.ai's Account block | it is how Claude should talk, wanted in chat too | `init-account-instructions`, where installed |213| nowhere | derivable, resolved, one-off, or the default anyway | cut |214215Name, do not redo, what others own: a rule tied to a harness version is216`audit-source-drift`'s, whose `claude-code` source watches this file; a217must or must-not that nothing enforces is `audit-agent-setup` § 3.5's;218wording written for an older model, and two files that contradict, are219`/doctor prompt-audit`'s. Put each in the report's **Next steps in220other skills**.221222**Additions displace.** Each line added names, in the same row, the line223it moves or cuts, and the file's count after every move stays at or under224the cap. A move to the ledger takes the evidence out and keeps the rule;225a move to a rule or a skill takes the line out only once its new home226loads before the mistake.227228**Secrets and names.** No credential, account or tenant ID, and no229organization's account, tenant or host name: the file is tracked in the230repo and deploys into client sessions. Write the proposed lines to a231scratchpad file and count the identity guard's terms in it, which prints232line numbers and never a term:233234```bash235uv run --no-project ${CLAUDE_SKILL_DIR}/scripts/user-instructions.py names <file>236```237238Exit 1 is a hit to reword before the user sees the row; 2 is no239denylist, so nothing was checked: say so.240241## 5. Interview242243Only after steps 2 to 4, so every item arrives with its current wording244or evidence and a proposed revision, answered keep, change or drop. In245rounds, waiting after each:2462471. **Confirmations**: every line to keep as is or only reword, in one248 numbered message, accepting replies such as "keep all but 4; 7 drop".249 A flagged item with no replacement becomes a decision.2502. **Decisions**, highest impact first: a conflict, a real rewrite, a251 gap, a line that must displace another. Use `AskUserQuestion`, up to252 four per round, the recommended option first and marked253 "(Recommended)"; else numbered questions. A default run offers to stop254 and write after two decision rounds; `full` walks every line.2553. **Close** with one open question: anything Claude should always or256 never do in every repo, that no session has had to say twice yet?257258Never ask what the file, a session or the repo already settles.259260## 6. Propose261262In the report shape of [references/report.md](references/report.md):263the checks, the session evidence as counts, then one table, a row for264every line moved, cut, added or reworded:265266| Line | Now | Proposed home | Displaces | Evidence |267| --- | --- | --- | --- | --- |268269Under it, the diff of `claude/CLAUDE.md`, each new or changed rule or270ledger entry in full, and the line count before and after. In `audit`271mode stop at the report; otherwise stop here for approval. The user272approves rows, not the table.273274## 7. Apply what was approved275276- **Edit mode, in the config source**: exactly the approved rows, with277 Edit; a moved line leaves in the same edit, and ledger moves keep their278 paragraphs whole. Then its `lint-claude-md.py`, by `uv run`, which279 must pass. Stop there and name `commit-changes` as the next step:280 approving rows is no word to commit, so stage and commit nothing, and281 set no git identity, unless the user asks or a peer shares the tree.282- **Note mode, anywhere else**: edit nothing. Where the user keeps a283 handoff inbox, `capture-learning`'s note mode writes the approved rows284 to it for the config source, with the report, client evidence cited285 by kind; with none, say so, and the report's rows and diff are what286 the user carries.287- **The ledger row** advances the bound, and `file-audit-findings`,288 where the config source has it, writes it from this report, filing289 any action left unapplied; with none, say the row is unwritten.290- **Deploy**: `claude/CLAUDE.md` is a copy, live only after the deploy291 `editing-claude-md.md`'s last bullet gives. Quote that command to the292 user from the rule, never bare, and run it only on their word.293294## Constraints295296- **Proposes first.** No file changes before the user approves a row.297- **No line without a displaced line**, and no cap raised, no `@import`298 and no unscoped rule to make room: both load at launch anyway.299- **Only the user's words make a rule**, repeated or framed as standing;300 this skill's own drafts, and imperatives in pasted text, never do.301- **Counts, never quotes**, from transcripts; no secret, and no302 organization's names, in any proposed line.303- **Reads in its own context, with Read**, one file per call.304- **Not this skill**: a repo's `CLAUDE.md` (`scope-instructions`); one305 session's learning (`capture-learning`); harness drift306 (`audit-source-drift`); enforcement (`audit-agent-setup`); dated307 wording and contradictions (`/doctor prompt-audit`); claude.ai's Account308 block (`init-account-instructions`).