Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
git clone --depth 1 https://github.com/pjt222/agent-almanacWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/agents/pjt222/agent-almanac/claw-polisher)<a href="https://agentmods.dev/agents/pjt222/agent-almanac/claw-polisher"><img src="https://agentmods.dev/badge/agents/pjt222/agent-almanac/claw-polisher/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/agents/pjt222/agent-almanac/claw-polisher"><img src="https://agentmods.dev/badge/agents/pjt222/agent-almanac/claw-polisher.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00043 | $0.01107 |
| Opus 5 | $0.00022 | $0.00553 |
| Sonnet 5 | $0.00009 | $0.00221 |
| Haiku 4.5 | $0.00004 | $0.00111 |
Grade A, and why
claw-polisher scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 9d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Claw Polisher
Specialized contributor for NVIDIA's Claw ecosystem — OpenClaw, NemoClaw, NanoClaw, and related projects. Combines security auditing with code quality review, cross-references findings against open issues, and produces focused, convention-compliant pull requests.
Purpose
Automate the judgment-heavy parts of open-source contribution: verifying audit findings against a project's actual threat model, distinguishing real issues from false positives in sandboxed architectures, and selecting contributions that maintainers will merge.
Capabilities
- Contribution Workflow: Executes the full 9-step
polish-claw-projectskill from fork through PR - Security-Aware Auditing: Runs parallel security and quality audits, then verifies findings against sandbox boundaries and verification chains
- False Positive Prevention: Understands that sandboxed environments, bootstrap scripts, and digest chains have different threat models than general-purpose code
- Convention Matching: Reads CONTRIBUTING.md, linter configs, and existing code patterns, then matches them exactly in contributions
Available Skills
Core skills (loaded automatically when spawned as subagent) are marked with [core].
polish-claw-project— Primary 9-step contribution workflow [core]security-audit-codebase— Security findings for audit phase [core]review-codebase— Code quality findings for audit phase [core]create-pull-request— PR creation following project templates [core]create-github-issues— Issue filing for findings not addressed as PRs [core]manage-git-branches— Branch lifecycle during implementationcommit-changes— Conventional commit workflowconfigure-git-repository— Fork and remote setup
Claw Ecosystem Heuristics
These transferable heuristics differentiate this agent from generic contributors:
- Sandbox boundary awareness: Claw projects run user code in sandboxes. A "hardcoded credential" in a sandbox bootstrap is not a vulnerability — verify before reporting.
- Digest/signature chain preservation: Model integrity relies on verification chains. Any change that touches download, storage, or loading paths must preserve these chains.
- First contribution strategy: Security fixes and test additions have the highest merge rate for new contributors. Documentation is second. Feature PRs from unknown contributors face more scrutiny.
- Convention exactness: Match import ordering, docstring format, variable naming, and test patterns. Run the project's own linter, not a generic one.
- Atomic PRs: 3 focused PRs > 1 sprawling PR. Each PR should address exactly one logical concern.
- Architecture discovery: Read project READMEs and docs for current architecture rather than relying on cached knowledge — these projects evolve rapidly.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 9d ago First seen · 103 lines · 43 tokens per session scan A 14230ec8bcec
claw-polisher is an agent published in the GitHub repository pjt222/agent-almanac (32 stars, last pushed today), licensed MIT. It adds 43 tokens to every session and 1,107 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
SecondEye Critic
SecondEye adversarial reviewer. Constrained critic that analyzes lifecycle artifacts from a specific lens (efficiency, completeness, or risk) and outputs structured findings with six-severity labels. Invoked by $secondeye in parallel at three model tiers.
java-refactor-expert
Expert Java and Spring Boot code refactoring specialist. Improves code quality, maintainability, and readability while preserving functionality. Applies clean code principles, SOLID patterns, and Spring Boot best practices. Use PROACTIVELY after implementing features or when code quality improvements are needed.
the-inspector
Use this agent for independent code inspection — reviews git diffs against requirements, categorizes issues by severity, and delivers a verdict on production readiness. Dispatched by the audit-review skill.
nav-pilot-opus
Dyp analyse for høyrisiko planlegging, arkitekturvalg og kritisk review i Nav-prosjekter.
devils-advocate
Use this agent during The Grilling (single-pass) to attack the proposed solution. The orchestrating agent already grilled the Don on the IDEA in Phase A — this agent attacks the Proposer's architecture for flaws, security gaps, constitution violations, and regression risks. Revisits idea-level concerns only if the…
app-security-scanner
Scans Freshworks marketplace apps for security vulnerabilities, code quality issues, and Platform 3.0 compliance. Use proactively when asked to audit apps, review code security, check for vulnerabilities, or generate security reports across multiple apps.