Skill v1.0.0

name: apply description: Apply one tokendoctor audit fix after showing the exact change. Moved to /tokendoctor fix <id>, which needs no model call. disable-model-invocation: true allowed-tools: Bash(tokendoctor-apply *) argument-hint: <finding-id> [--scope user|project|local] metadata: version: "1.0.0"

Apply the fix for the audit finding named in $ARGUMENTS. First say in one line that /tokendoctor fix <id> does the same with no model call and that this command goes away in the next release.

1. Run tokendoctor-apply $ARGUMENTS with no --write. It prints which files would change and a

short -/+ summary, and changes nothing.

1. Show the user that output: the file path, the lines removed and added, and the scope (user

settings affect every project; project settings are shared with the team through git; local settings are personal to this project).

1. Ask for confirmation. Do not proceed on silence or an unrelated reply.
2. After a clear yes, run the same command with --write. It copies an existing file to

<file>.tokendoctor-<timestamp>.bak before writing, and never overwrites an existing agent file.

1. Report what was written and the backup path. Most settings apply to the running session, but

say that a new session is the surest way to confirm, and that /status lists loaded files.

Never edit settings files by hand from this skill, and never pass --write before the user has seen the dry run.