agentleFS
Sign inStart free
arndvsarndvs/ctrlshft/skills/press1-check/SKILL.md

press1-check

Skill0 starsChanged 6 months agoMITView on GitHub

Use when Claude Code sessions had many manual approval ("press 1") prompts or when auditing hook permissions; identifies which Bash commands required approval.

  • Deletes or force-pushes
  • Installs packages
  • Commits and pushes

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 5 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

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

    If running interactively (human present), output "Read Press1 Check skill." to acknowledge.

    line 8
  2. Distinctive:

    After sessions with many "press 1 to approve" prompts

    line 16
  3. Distinctive:

    When the user says "permission audit" or "press1-check"

    line 18
  4. Distinctive:

    Before proposing new hooks that add Bash commands

    line 19
  5. Very specific:

    Run python3 <skill-path>/audit-permissions.py with any arguments the user provided.

    line 58
  6. Very specific:

    If LOW-risk suggestions appear, add them to ~/.claude/settings.json (read-only commands are safe).

    line 60
  7. Distinctive:

    Do NOT auto-add MEDIUM or HIGH risk commands without explicit approval.

    line 61
  8. Very specific:

    After editing ~/.claude/settings.json, re-run the audit and report before/after counts so the user sees the coverage delta.

    line 62
  9. Distinctive:

    Never auto-add MEDIUM or HIGH risk commands

    line 79

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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

/press1-check — Permission Audit

If running interactively (human present), output "Read Press1 Check skill." to acknowledge. If running with --dangerously-skip-permissions (AFK/unattended), skip acknowledgement and proceed directly.

Audit which Bash commands triggered manual approval prompts in Claude Code sessions.

When to invoke

  • After sessions with many "press 1 to approve" prompts
  • Periodically (every 5-10 sessions) to keep the allow-list current
  • When the user says "permission audit" or "press1-check"
  • Before proposing new hooks that add Bash commands

How it runs

Two paths:

  • Manual: type /press1-check to force a re-audit. Default mode is since the last run (state-tracked at ~/.claude/state/press1-check.json), so the typical run only surfaces new ground.
  • Auto (optional): the script supports --auto-stop-hook mode designed to be wired as a Claude Code Stop hook. 6-hour cooldown. When LOW-risk additions are found, a one-shot summary surfaces in the next session-start priority snapshot.

Usage

# Since the last run (default, state-tracked)
python3 skills/press1-check/audit-permissions.py

# Last N days across all project dirs
python3 skills/press1-check/audit-permissions.py --days 7

# Only the single most recent session
python3 skills/press1-check/audit-permissions.py --latest-session

# All sessions from the last 24h
python3 skills/press1-check/audit-permissions.py --all-recent

# All sessions since a date
python3 skills/press1-check/audit-permissions.py --since 2026-04-10

# Specific session (prefix match OK)
python3 skills/press1-check/audit-permissions.py <session-id>

Steps (when invoking via /press1-check)

  1. Run python3 <skill-path>/audit-permissions.py with any arguments the user provided. Default audits since ~/.claude/state/press1-check.json#last_run_ts (bootstraps to last 3 days when state is missing).
  2. Display the output to the user exactly as printed (includes color-coded risk levels).
  3. If LOW-risk suggestions appear, add them to ~/.claude/settings.json (read-only commands are safe). Skip env-var-prefix suggestions like Bash(WT=*) — they don't generalize.
  4. Do NOT auto-add MEDIUM or HIGH risk commands without explicit approval.
  5. After editing ~/.claude/settings.json, re-run the audit and report before/after counts so the user sees the coverage delta.
  6. Each successful interactive run updates the state file — the next run is automatically scoped to "since now."

Risk Classification

RiskColorPolicyExample
HIGHRedKeep gated — destructive or hard to reverserm, sudo, git reset --hard, DROP TABLE
MEDIUMYellowReview case-by-case — side effects outside local repogh pr create, curl, docker, pip install
LOWGreenSafe to allow-list — read-only or local-onlycat, ls, grep, wc

Constraints

  • Never auto-add MEDIUM or HIGH risk commands
  • The script never auto-applies changes — it surfaces and proposes
  • State-tracked incremental runs prevent audit fatigue
  • Subagent awareness: detects */subagents/*.jsonl and tags findings

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.

Reviews can't be read right now.

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.

Check your setup, free