agentleFS
ivyivy/dotfiles/.claude/skills/allow/SKILL.md

allow

Skill15 starsChanged 13 months agoISCView on GitHub

Add a tool pattern to the global Claude Code permissions allow list in bin/sync-claude-settings

  • Commits and pushes

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Add tool patterns to the global Claude Code permissions managed by bin/sync-claude-settings.

    line 18
  2. Very specific:

    Run git diff bin/sync-claude-settings and verify:

    line 91

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Allow Tool

Add tool patterns to the global Claude Code permissions managed by bin/sync-claude-settings.

Arguments

/allow git check-ignore
/allow limactl
/allow /think
/allow docs.rs

Instructions

1. Parse the pattern

Infer the full permission string from the argument:

InputResulting patternWhy
git check-ignoreBash(git check-ignore:*)bare words → Bash
limactlBash(limactl:*)bare word → Bash
/thinkSkill(think:*)leading slash → Skill
/commitSkill(commit:*)leading slash → Skill
docs.rsWebFetch(domain:docs.rs)looks like a domain → WebFetch
raw.githubusercontent.comWebFetch(domain:raw.githubusercontent.com)domain → WebFetch
Bash(make:*)Bash(make:*)already formatted
Skill(think:*)Skill(think:*)already formatted
WebFetch(domain:docs.rs)WebFetch(domain:docs.rs)already formatted

Heuristics:

  • Starts with / → Skill(<name>:*)
  • Contains a . and looks like a domain (no spaces) → WebFetch(domain:<input>)
  • Already wrapped in Bash(...), Skill(...), WebFetch(...), Read(...) → pass through
  • Everything else → Bash(<input>:*)

2. Decision framework

A command belongs on the allow list based on its side-effect profile:

Allow (no confirmation needed):

  • Pure reads — queries state, never mutates it (ps, stat, file, which, uname, dig)
  • Pure transforms — reads stdin/files, writes only to stdout (sort, uniq, cut, jq, base64, shasum)
  • Bounded benign writes — side effects are trivial and easily reversed (mkdir, touch, mktemp, rmdir empty dirs only)

Reject (warn, require explicit confirmation): A command does NOT belong if it can:

  • Write to arbitrary files — tee, cp, mv, output redirection
  • Execute arbitrary subcommands — xargs, sh, bash, eval, python, ruby, perl, node
  • Publish or transmit data — curl, wget, ssh, scp, rsync, git push
  • Destroy state — rm, kill, killall
  • Escalate privileges — sudo, doas
  • Mutate permissions or identity — chmod, chown
  • Access or leak secrets — gpg, security, keychain, op, env, printenv, history
  • Mutate cloud infra — aws, gcloud, az
  • Mutate global config — git config (can write ~/.gitconfig via --global)

Warn and suggest narrowing for overly broad patterns:

  • Bash(*) — universal wildcard
  • Multi-subcommand tools where we already scope by subcommand: Bash(git:*), Bash(npm:*), Bash(gh:*)
  • WebFetch(domain:*) — wildcard domain
  • Read(...) / Edit(...) with broad paths — can expose or modify sensitive files

For reject and narrowing cases: explain which property is violated, suggest a safer alternative if possible, and ask the user to confirm. If they confirm, proceed.

3. Check for duplicates

Grep bin/sync-claude-settings for the exact resolved pattern. If it already exists, tell the user and stop.

4. Add the pattern

Edit bin/sync-claude-settings, inserting the new pattern into the set_permissions() function's allow array in alphabetical order among its peers.

5. Review the edit

Run git diff bin/sync-claude-settings and verify:

  • Exactly one line was added
  • The jq heredoc is still valid (no unclosed quotes)
  • The pattern is in the correct alphabetical position

6. Apply and commit

bin/sync-claude-settings

Then use /commit to commit bin/sync-claude-settings.

Examples

/allow git check-ignore    → adds "Bash(git check-ignore:*)", applies, commits
/allow limactl             → adds "Bash(limactl:*)", applies, commits
/allow /think              → adds "Skill(think:*)", applies, commits
/allow docs.rs             → adds "WebFetch(domain:docs.rs)", applies, commits
/allow git push            → REJECTS: publishes to remote. Proceeds only if user confirms.
/allow printenv            → REJECTS: can leak secrets from environment. Proceeds only if user confirms.
/allow npm                 → NARROWS: too broad, suggests npm run, npm test, etc. Proceeds only if user confirms.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.