summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorThomas Vanbesien <tvanbesi@proton.me>2026-06-12 11:50:40 +0200
committerThomas Vanbesien <tvanbesi@proton.me>2026-06-12 11:52:48 +0200
commit812025bee178a4959bdca10b3e6c8d84ca3580d9 (patch)
tree7692018ba6045fa1161909237c05a05f67bb7f80
parent29cf89ff5b26dd7443b5fcd5d9bd6f49e46b7b15 (diff)
downloaddotfiles-812025bee178a4959bdca10b3e6c8d84ca3580d9.tar.gz
dotfiles-812025bee178a4959bdca10b3e6c8d84ca3580d9.zip
misc(claude): edit file format
-rw-r--r--.claude/CLAUDE.md10
1 files changed, 3 insertions, 7 deletions
diff --git a/.claude/CLAUDE.md b/.claude/CLAUDE.md
index 172c5e5..13c5916 100644
--- a/.claude/CLAUDE.md
+++ b/.claude/CLAUDE.md
@@ -6,7 +6,7 @@
# CLAUDE.md memory
-- Create project memory CLAUDE.md at `./.claude/CLAUDE.md,` not `./CLAUDE.md`. Do not commit it to git.
+- Create project memory CLAUDE.md at `./.claude/CLAUDE.md,` not `./CLAUDE.md`. Do not commit it to git. Verify it is there once when starting, because I replace it with a symlink `./.claude/CLAUDE.md` → `~/.claude/projects/PROJECT_DIR/CLAUDE.md` so that I can put the original in a synced folder and use it on multiple devices.
# Auto-memory
@@ -37,12 +37,8 @@
# Mode
-- Default to diagnostic mode: investigate, explain, recommend. Don't reach for Edit/Write unless I explicitly ask for a
- change ("do X", "make the change", "fix it", "apply that"). If ambiguous, ask.
+- Default to diagnostic mode: investigate, explain, recommend. Don't reach for Edit/Write unless I explicitly ask for a change ("do X", "make the change", "fix it", "apply that"). If ambiguous, ask.
# Running scripts
-- When you write a script that does real work (data transformation, transfers, mutations, anything with side effects),
- output it for me to run myself — don't execute it. Running is mechanically trivial; what matters is that I stay in
- control of when its effects happen. Ad-hoc read-only commands (greps, status checks, inspections) are fine to run
- directly.
+- When you write a script that does real work (data transformation, transfers, mutations, anything with side effects), output it for me to run myself — don't execute it. Running is mechanically trivial; what matters is that I stay in control of when its effects happen. Ad-hoc read-only commands (greps, status checks, inspections) are fine to run directly.