Skills DirectorySkills Directory
SkillsLearnSecurityCategoriesDocsBlogPro
Sign InSubmit Skill
Skills Directory

Security-tested agent skills for Claude, coding agents, and AI workflows.

Directory

  • Browse Skills
  • All Skills A–Z
  • Claude Skills
  • Claude Code Skills
  • Agent Skills
  • Categories
  • Authors
  • Submit a Skill

Learn

  • Learn Hub
  • Install Claude Skills
  • Write SKILL.md
  • Skills vs MCP
  • Directories Compared

Security

  • Security
  • Methodology
  • Secure Claude Skills
  • Security Badges
  • Chrome Extension
  • Skill Manager

Company

  • About
  • Community
  • Blog
  • API Docs
  • Advertise

2026 Skills Directory. All rights reserved.

ProTermsPrivacyRefunds
Back to skills

Pawl

DSecurity

Deterministic gates for agent harnesses. Use when a call is denied, prompted or blocked with a [PAWL blast|egress|prose|budget|git|pin|teardown|tamper|poll|loop|no-op|reread|fence|creds|verify] or [IDLE TASK] reason, or before sending a message, discarding git work, installing from a URL, waiting on a background task, fixing a bug with a reproducer, or growing an always-loaded prompt file.

2 stars
0 votes
0 copies
0 views
Added 10/6/2026
ai-agentspythongoshellbashterraformgitdatabase

Works with

claude codecli

Security Analysis

D42/100
criticalPipes output to a shell interpreter
criticalDownloads and executes remote scripts — classic supply chain attack
mediumInstalls packages at runtime which could introduce malicious dependencies

Pro scans all 20 files and shows the line behind each finding

Scanned 10/7/2026

$npx -y skills add ulukaya/pawl --skill pawl --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Pawl?

Add the live security badge to your README — it updates automatically with every re-scan.

Security grade badge for Pawl
[![Security: D — Skills Directory](https://www.skillsdirectory.com/api/skills/ulukaya-pawl/badge)](https://www.skillsdirectory.com/skills/ulukaya-pawl)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
Files
SKILL.md
---
name: pawl
description: "Deterministic gates for agent harnesses. Use when a call is denied, prompted or blocked with a [PAWL blast|egress|prose|budget|git|pin|teardown|tamper|poll|loop|no-op|reread|fence|creds|verify] or [IDLE TASK] reason, or before sending a message, discarding git work, installing from a URL, waiting on a background task, fixing a bug with a reproducer, or growing an always-loaded prompt file."
---

# pawl

`<root>` is the plugin root, two directories above this file. Pieces live
under `<root>/pieces/<name>/`; every hook runs `<root>/hooks/pawl.py`. State
lives under `PAWL_DATA` (default `~/.pawl`).

## Denial prefixes

-   `[PAWL egress]`, `[PAWL prose]`, `[PAWL budget]` -> `send-gates.md`
-   `[PAWL blast]` -> `blast-radius-guard.md`
-   `[PAWL git]` -> `destructive-git-guard.md`
-   `[PAWL poll]` -> `poll-loop-guard.md`
-   `[PAWL loop]` -> `oscillation-breaker.md`
-   `[PAWL no-op]` -> `noop-edit-guard.md`
-   `[PAWL reread]` -> `reread-guard.md`
-   `[PAWL fence]` -> `conversation-fence.md`
-   `[PAWL creds]` -> `credential-fence.md`
-   `[PAWL pin]` -> `install-pin-guard.md`
-   `[PAWL teardown]` -> `teardown-guard.md`
-   `[PAWL tamper]` -> `check-tamper-guard.md`
-   `[PAWL verify]` -> `verify-gate.md`
-   `[IDLE TASK]` -> `idle-task-gate.md`

## Routing

Each arrow names the page to read under `references/` beside this file.

-   Send (chat, mail, social post): gate `send`, egress then prose then
    budget -> `send-gates.md`
-   A deletion outside the workspace, or of home, root or a drive, in a
    command or in what it runs: gate `blast` denies or asks ->
    `blast-radius-guard.md`
-   Destructive `git` in a protected repo, `git commit -n`, a worktree on
    tmpfs, a force push, `git branch -D`: gate `git` asks; by hand
    `destructive_git_guard.py check` -> `destructive-git-guard.md`
-   Poll loop, `tail -f`, `watch`, `sleep` over 600 s: gate `poll` asks; by
    hand `poll_loop_guard.py classify` -> `poll-loop-guard.md`
-   Turn end with background waits still running: gate `idle` blocks once ->
    `idle-task-gate.md`
-   Changed configured project without a passing receipt: gate `verify`
    reminds once -> `verify-gate.md`
-   Same tool call 3 times, or 2-3 calls alternating: gate `loop` asks ->
    `oscillation-breaker.md`
-   Edit whose replacement equals its target, or a write of the bytes
    already on disk: gate `noop` denies -> `noop-edit-guard.md`
-   Invisible characters in a write: gate `zero-width` strips them ->
    `zero-width-sanitizer.md`
-   Read-only shell command: gate `readonly` approves; by hand
    `readonly_pass.py check` -> `readonly-pass.md`
-   Re-reading the own transcript, a `SKILL.md` or a memory file: gate
    `reread` denies past the limit -> `reread-guard.md`
-   Reading another conversation's files: gate `fence` asks ->
    `conversation-fence.md`
-   Reading browser cookies, saved passwords or a keychain: gate `creds`
    asks -> `credential-fence.md`
-   Installing from a branch URL, a global npm install without a version,
    `curl | sh`: gate `pin` asks -> `install-pin-guard.md`
-   Dropping or emptying a remote database, a migration reset, `terraform
    destroy`, `kubectl delete namespace`: gate `teardown` asks ->
    `teardown-guard.md`
-   Skipping or `.only`-ing a test, `assert True`, `|| true` in CI,
    turning hooks off: gate `tamper` asks -> `check-tamper-guard.md`
-   Which gates fired this week: `report.py --days 7` -> `report.md`
-   Text a human will read: `prose_gate.py --plane chat|deliverable` ->
    `prose-gate.md`
-   Outbound text with paths, tokens, hostnames: `egress_firewall.py check`
    -> `egress-firewall.md`
-   Sends per channel per day: `send_budget.py status|check|spend` ->
    `send-budget.md`
-   A count that may only go down: `ratchet.py check|update|show` ->
    `ratchet.md`
-   Scheduled job failing repeatedly: `breaker.py run <name> -- <cmd>` ->
    `circuit-breaker.md`
-   Bug fix needing a failing reproducer: `repro_fence.py red|fence|both` ->
    `repro-fence.md`
-   A test that may not test anything: `bite_check.py check` ->
    `bite-check.md`
-   Always-loaded files growing: `prompt_budget.py check|report --config
    budget.json` -> `prompt-budget.md`

## Contracts

-   CLIs exit 0 pass, 1 fail, 2 internal error or missing input (piece pages
    list extras: breaker 3 = skipped).
-   A gate that asks: Antigravity `force_ask`, Claude Code `ask`. A Codex
    PreToolUse hook cannot ask, so there it denies with the same reason; tell
    the user.
-   The egress firewall, git and poll guards fail closed; every other gate
    fails open. A run past its watchdog asks on a shell or write call
    (Codex denies): tell the user the call was too slow to check.
-   Skip gates for a session: `PAWL_DISABLE=git,poll,send,...` (any gate
    name, or `egress`, `prose`, `budget`); `<root>/hooks/pawl.py gates` lists
    them.
-   Never edit a piece to make a gate pass. Fix the draft, command, count, or
    reason.

## Verify

```bash
python3 -B <root>/run_tests.py
python3 -B <root>/check_portable.py
```

Expected: one `OK` line per suite, then `portable: clean`, exit 0.

Attribution

ulukayaulukaya
View sourceSee grades on GitHubMore from ulukaya →
SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments (0)

No comments yet. Be the first to comment!

SSkills DirectorySkills Directory

Ship a skill? Prove it's safe.

Free 120-pattern security scan, letter grade, and an embeddable README badge.

Submit a skill

Related Skills

Caveman

Terse caveman voice: answer first, fluff gone, every technical fact kept. Use for /caveman, "caveman mode", "talk like caveman", "be brief", "less tokens". Stays on until "stop caveman" or "normal mode".

1100021 votes

Hyperplan

Adversarial multi-agent planning skill. Self-orchestrates 5 hostile category members (unspecified-low, unspecified-high, deep, ultrabrain, artistry) via team-mode for ruthless cross-critique debate, distills only the defensible insights, then MANDATORILY hands the distilled insight bundle to the `plan` agent for executable plan formalization. Use when planning needs maximum rigor and surfacing of weak assumptions, blind spots, and over-engineering. Triggers: 'hyperplan', 'hpp', '/hyperplan', ...

698621 votes

Writing Skills

Create and manage Claude Code skills in HASH repository following Anthropic best practices. Use when creating new skills, modifying skill-rules.json, understanding trigger patterns, working with hooks, debugging skill activation, or implementing progressive disclosure. Covers skill structure, YAML frontmatter, trigger types (keywords, intent patterns), UserPromptSubmit hook, and the 500-line rule. Includes validation and debugging with SKILL_DEBUG. Examples include rust-error-stack, cargo-dep...

3931 votes

Mcp Code Execution

Routes multi-tool workflows through MCP servers for large datasets and pipelines. Use when Bash tool overhead is limiting throughput on data-heavy tasks.

3421 votes

catchup

Recovers the conversation and failed tool calls of a previous Codex, Amp, Claude Code, Antigravity, Cline, Copilot CLI, Cursor, DeepSeek Harness, Grok Build, Kimi, OpenCode, Pi Agent, or ZCode session. Use when the user says "catch up", "what did the last session do", "get me up to speed", "I switched agents", asks to recover/summarize a previous session before continuing, or asks to diagnose or report a catchup failure. Do NOT use for the current conversation, git history, or any non-agent log.

741 votes
View all in ai-agents →