permissions-tune
Use when permission prompts get in the way: fewer permission prompts, pre-approve read-only commands, why was that denied, allowlist commands I keep approving, make auto mode my default, stop asking me for git log. Proposes auto mode as the default mode and exact allow rules for read-only calls that keep prompting, from transcript evidence.
- License 许可
- GPL-3.0
- 最近更新
- 2026-11-13
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- permissions-tune
- description
- Use when permission prompts get in the way: fewer permission prompts, pre-approve read-only commands, why was that denied, allowlist commands I keep approving, make auto mode my default, stop asking me for git log. Proposes auto mode as the default mode and exact allow rules for read-only calls that keep prompting, from transcript evidence.
03 SKILL.md 正文
Permissions tune
Every prompt for a read-only command costs a decision that changes nothing. The built-in
/fewer-permission-prompts writes wildcard rules into project settings; this skill writes exact
rules for the observed calls into the local settings, because a wildcard also approves flags nobody
saw. Propose two things:
auto mode as the default, and exact allow rules for read-only calls that keep prompting. Both
widen what runs unasked, so each is its own question naming every change. Shared reading and apply
rules are in the usage skill. A prompt saves no tokens: savingTokens is 0, and savingNote
says the prompts removed ("~9 prompts per session").
1. Auto mode
Read permissions.defaultMode by key from user, project and local settings. Healthy: user says
"auto" and nothing overrides it. A project value other than "auto" wins over the user's: say
so. Skip with one line when managed policy sets defaultMode or any file has disableAutoMode: "disable". Otherwise propose "permissions": {"defaultMode": "auto"} in ~/.claude/settings.json
only, because "auto" in project settings is ignored. It applies to every project.
2. Allow rules
The ledger records uses, not prompts, so read the ~50 newest transcripts.
- Collect
userentries withtoolDenialKind(user-rejected,permission-rule,automode-*);interruptedandcancelledare aborts. Followtool_use_idback to thetool_usefor name and input. Older transcripts without the field: match the denial text, label counts unverified, never formcp__*(the server writes that text). - Rank. Key Bash on command plus first subcommand (
git log), MCP on the fullmcp__<server>__<tool>. Drop under ~3 uses, keep the top ~20. - Keep only read-only. Drop: interpreters and runners (
python,node,bun,bash,npx,bunx,uvx); wildcards of task runners (npm run *,make *,just *; exactBash(bun run typecheck)is fine);curl,wget,git fetch,git pull, anygh api,docker exec,kubectl exec,sudo;find -execor-delete; any-c key=value,--exec-path,--upload-pack,VAR=xprefix, pipe or redirection; anything under an existingdenyoraskrule; commands Claude Code already allows (ls,cat,git status,rg, most read-only git and gh), because a denial there came from a rule or the classifier. - Exact rules only:
Bash(git log --oneline -20), neverBash(git log *), because read-only subcommands have writing flags (git log --output=<file>,git branch -D) and transcript strings can be steered by prompt injection: an exact rule approves only what was observed. MCP: one full tool name per rule, nevermcp__x__get_*. - Mostly
user-rejectedmeans the person said no: include the rule only with that stated.
Destination: permissions.allow in .claude/settings.local.json, never user scope, because the
evidence spans every project. Deduplicate; never touch deny or ask.
Findings
One finding per rule and one for the mode: change.kind: "edit", target the settings file, the
rule in detail, verdict: "fix", question: "permissions". Asked alone: a table | # | Rule | Count | Denial kinds | Why read-only |, then what was skipped and why ("dropped git push:
writes; dropped cat: already allowed"). Ask once; apply only on yes; then say the file, the rules
added, and how to remove them.
Red flags
- When a candidate is a wildcard, narrow it to the observed command or drop it, because a wildcard approves flags nobody saw.
- When the command was denied by a
denyrule, skip it, because the person decided already.
Say it in the reply
The table, what was skipped, and every rule and mode change awaiting a yes.
04 同仓库更多 Skills
worktrunk
Use when working with git worktrees or wt (worktrunk): creating, switching, merging or removing a branch workspace, configuring wt (project or personal config, workflow hooks around switch, commit and merge, commit-message generation, hook approvals), or a wt command that failed. Triggers: make a worktree for this branch, wt merge, wt remove, configure wt, .config/wt.toml, run tests before merge, install deps when a worktree is created, wt hook, wt switch does not cd, nastav wt, wt nefunguje.
web-ladder
Use when a task needs something the internet holds: look something up, search the web, find which page answers a question, read a page, check a library's current docs, a page that needs a browser or shows a captcha. Runs one research loop: searxng search, keywords, cases, search again up to three times, Context7 for a library, Tavily when the keywords will not come, one page in a browser, the person on a wall. Triggers: look this up, search the web, find out online, read this page, check the docs, the page is blocked, captcha, najdi na webu, otevri tu stranku. A multi-round question goes to the qsm-web:web-research agent; one tool's commands are its own skill (searxng, browserbox, opensrc).
searxng
Use when running a web search through the person's own SearXNG instance with qsm-web searxng: choosing a search profile or engines, the instance, time range or language, reading its JSON, a suspended engine. Triggers: search with searxng, qsm-web searxng, search profile, niche search, academic search, IT search, vyhledavaci profil, searxng is down, which engines. The research loop around it is the web-ladder skill.
opensrc
Use when the source code of an open-source library or repository is needed on disk to read: the docs are silent, the behaviour must be checked in code, the version the project pins. Runs qsm-web opensrc get with a package spec or a repository address. Triggers: get the source of this package, clone this library, read the library's code, qsm-web opensrc, stahni zdrojak knihovny. Library docs are the Context7 skill; the research loop is the web-ladder skill.
browserbox
Use when one page needs a real browser through qsm-web browserbox: content built by script, a page that blocks plain fetching, a captcha to solve or hand to the person, a scripted sequence of clicks, a login the person is present for. Triggers: open it in a browser, the page is empty without JavaScript, captcha, take over the browser, run these steps on the page, log in to this site, browserbox. When to reach for a browser at all is the web-ladder skill.
usage
Use when the user wants a review of how they use Claude Code, or setup hands over after a first install: review my Claude Code usage, audit my setup, what does my setup cost, where did my tokens go, clean up my Claude Code, usage report, zkontroluj moje pouzivani. One walkthrough of four stages (machine, work, habits, extensions), questions after each, then one plan. A question about one topic (which hooks are slow) goes to that topic's skill; this session's tokens alone is `/qsm-usage`.