using-scrimlisted

# Using Scrim Scrim is a context firewall. Sensitive content is replaced with stable tokens of the form `⟦scrim:<class>:<id>⟧` before it ever enters your context. You can reason about and edit that content normally — when you write a file back, Scrim's PreToolUse hook de-tokenizes the tokens to the real values on disk. ## When to use Scrim tools Prefer Scrim's tools over native `Read` and `Bash` whenever the target could contain credentials or personal data: - **`safe_read`** — for any config-like file: `.env*`, `*.tfvars`, `*.pem`, `config.{json,yml,yaml,toml}`, `settings.py`, files under `secrets/**`, seed/fixture data with personal info. - **`safe_grep`** — for searches across paths that may include the above. - **`safe_shell`** — for commands whose output frequently contains secrets: `env`, `printenv`, `kubectl get secret -o yaml`, `docker inspect`, `git remote -v`, anything that prints a connection string. - **`safe_write_token`** — when you need to MODIFY a value behind a token (rotating a password, replacing a secret, changing a URL inside a tokenized connection string). The token slug stays the same; only the underlying value changes. Call this BEFORE the `Write`/`Edit` that should land the new value — every place that token appears in a file you write back will get the new value. Use native `Read`/`Bash` for everything else (source code, build output, etc.). ## How tokens behave A token like `⟦scrim:db_password:a1b2c3⟧` is opaque but stable within a session. T