Untrack memory, state goals in own terms
This commit is contained in:
+1
-3
@@ -1,5 +1,5 @@
|
||||
# Allowlist: Claude Code writes credentials (.credentials.json), full session
|
||||
# transcripts (projects/, history.jsonl) and shell env dumps (shell-snapshots/)
|
||||
# transcripts (projects/, history.jsonl), memory and shell env dumps
|
||||
# into this directory, and each release adds more. Ignore everything, then
|
||||
# un-ignore authored config only.
|
||||
*
|
||||
@@ -14,8 +14,6 @@
|
||||
!agent-shared/**
|
||||
!agents/
|
||||
!agents/*.md
|
||||
!memory/
|
||||
!memory/*.md
|
||||
!principles/
|
||||
!principles/*.md
|
||||
!skills/
|
||||
|
||||
@@ -1,20 +1,20 @@
|
||||
# Claude Code configuration
|
||||
|
||||
The authored half of `~/.claude`: global instructions, agents, skills, principles, memory and settings. Publishes nothing.
|
||||
My personal `~/.claude`, version controlled. Instructions, agents, skills, principles and settings.
|
||||
|
||||
## Goals
|
||||
|
||||
1. **No secret and no transcript ever reaches a commit.** `.gitignore` is an allowlist: everything is ignored, and a file is tracked only by being named. A new Claude Code release dropping a token file next to these is untracked by default.
|
||||
2. **A clone restores the configuration on a new machine.** Anything needed to work the way these instructions describe is tracked; anything the tool regenerates is not.
|
||||
3. **The history explains why an instruction changed.** One behaviour change per commit, message saying what it now does.
|
||||
1. Share this configuration between my computers.
|
||||
2. Keep it backed up.
|
||||
3. Leak nothing — no credentials, no session transcripts, no memory.
|
||||
|
||||
## Audience
|
||||
|
||||
The owner of this machine, and the Claude Code sessions that read these files. Both may rely on everything tracked here being hand-written and safe to share — no credentials, no session data, no machine-generated state.
|
||||
Me, on my own machines. Private repo, nobody else depends on it.
|
||||
|
||||
## Tracking a new file
|
||||
|
||||
Add it to `.gitignore` as a negation, then `git check-ignore -v <path>` to confirm. Re-including a file needs its parent directories un-ignored too:
|
||||
`.gitignore` is an allowlist: everything is ignored, and a file is tracked only by being named. Add it as a negation, with its parent directories un-ignored too, then confirm with `git check-ignore -v <path>`:
|
||||
|
||||
```
|
||||
!commands/
|
||||
|
||||
@@ -1,2 +0,0 @@
|
||||
- [Run tests via Docker](feedback_run_tests.md) — Always use Docker-based test commands from README, not php artisan test
|
||||
- [Memory location preference](feedback_memory_location.md) — Store memories at user level (~/.claude/memory/), not per-project
|
||||
@@ -1,11 +0,0 @@
|
||||
---
|
||||
name: Store memory at user level
|
||||
description: User prefers memories stored at ~/.claude/memory, not in per-project memory directories
|
||||
type: feedback
|
||||
---
|
||||
|
||||
Store memories at the user level (~/.claude/memory/), not in per-project memory directories.
|
||||
|
||||
**Why:** User explicitly requested this — they want memories to be available across all projects, not scoped to individual ones.
|
||||
|
||||
**How to apply:** Always write memory files to ~/.claude/memory/ and maintain MEMORY.md there.
|
||||
@@ -1,11 +0,0 @@
|
||||
---
|
||||
name: Run tests via Docker
|
||||
description: Tests must be run using Docker as described in README.md, not via php artisan test directly
|
||||
type: feedback
|
||||
---
|
||||
|
||||
Always run tests using the Docker-based method described in README.md, not via `php artisan test` directly.
|
||||
|
||||
**Why:** The user corrected this — projects use Docker for test execution and `php artisan test` is not the correct way to run tests.
|
||||
|
||||
**How to apply:** Check the project's README.md for the exact test commands. Typically involves `docker compose run --rm tests` from the tests directory.
|
||||
Reference in New Issue
Block a user