Versioned security assessment

Report ID: SA-D71C7417

7/12/2026, 2:28:09 PM

cookbook-audit security assessment v13

Skill Security Certification Report

Audit History
Audit model: codex Latest published report
Skill name
cookbook-audit
Version
v13
Maintainer
anthropics
Source
github.com/anthropics/claude-cookbooks/tree/main/.claude/skills/cookbook-audit
Coverage
4 Files scanned · 962 Lines analyzed
Policy version
Unavailable

Highest confirmed finding severity

Medium

2 confirmed security findings require attention.

Installation context

Check the current Skill page

This page summarizes report evidence only. The Skill page provides the canonical install advisory.

Open current Skill page

This report does not block or authorize the manifest or ZIP.

Most static alerts are false positives caused by Markdown code formatting, documentation examples, placeholders, and ordinary validation text. Confirmed risks include shell command injection through interpolated paths, external process execution, unpinned runtime packages, and missing prompt-injection boundaries for reviewed notebooks.

Report position

Latest published report

Latest refers to the report sequence, not to artifact currentness.

Audit attestation

Not attestable

The required immutable binding is incomplete.

Human verification

Not verified

No human verification is recorded for this report.

Coverage

4 Files scanned · 962 Lines analyzed

7 items shown for review

Limitations

This report does not claim runtime or sandbox execution and does not prove the absence of side effects.

Evidence chain

Follow the evidence from source binding to the install contract. Available evidence supports verification; it is not a safety guarantee.

  1. Source

    Commit and path bound

  2. Artifact

    Content and tree hashes bound

  3. Audit

    Complete

  4. Install contract

    Open manifest to verify

    Open manifest

Capabilities observed

Observed means this report recorded supporting evidence. Not recorded does not prove that a capability is absent.

Contains scripts

May execute code included with the Skill.

Not recorded by this audit

Network access

May connect to external services.

Observed in 2 evidence locations

Filesystem access

May read or write local files.

Observed in 1 evidence location

Env variables

May read values from the process environment.

Observed in 14 evidence locations

External commands

May invoke commands or programs outside the Skill.

Observed in 19 evidence locations

Capability review items (5)
High
validate_notebook.py:82
Python subprocess.run
subprocess.run(cmd, capture_output=True, text=True, check=True) # noqa: S603
The validator launches uv, Jupyter, and nbconvert as external processes. Arguments avoid shell injection, but runtime package resolution and executable invocation remain real side effects.
High
validate_notebook.py:128
Python subprocess.run
result = subprocess.run( # noqa: S603
This subprocess executes sh -c with a command assembled from notebook and project paths. A single quote in a supplied path can escape quoting and inject shell commands.
High
validate_notebook.py:119
xargs command (can execute arbitrary commands)
f"echo '{notebook_abs}' | tr '\\n' '\\0' | xargs -0 uvx --from detect-secrets detect-secrets-hook --
The baseline branch interpolates notebook_abs inside single quotes and leaves baseline and plugin paths unquoted. Crafted paths can alter the sh -c command.
High
validate_notebook.py:125
xargs command (can execute arbitrary commands)
f"echo '{notebook_abs}' | tr '\\n' '\\0' | xargs -0 uvx --from detect-secrets detect-secrets-hook",
The fallback branch embeds the user-selected notebook path in a shell string. A path containing a single quote permits shell command injection.
Medium
SKILL.md:26
Ruby/shell backtick execution
3. **Run automated checks**: Use `python3 validate_notebook.py <path>` to catch technical issues and
This is an explicit instruction to execute validate_notebook.py against a user-selected path. The command launches the validator and its external dependencies.

Risk findings

Confirmed security concerns are separated from items that still need review.

Confirmed security concerns (2)

RISK-001 Medium
validate_notebook.py:66-82validate_notebook.py:114-133
Unpinned Runtime Package Execution
The validator runs nbconvert and detect-secrets through uv without version pins or hashes. Missing packages may be downloaded and executed during an audit.
Both command definitions name packages without versions, and the uv and uvx workflows resolve executable environments. This creates a reproducibility and supply-chain risk.
RISK-002 Medium
SKILL.md:30-36
Untrusted Notebook Instructions Lack Isolation
The workflow directs the agent to read notebook-derived Markdown but does not say to ignore embedded instructions. A crafted notebook can attempt prompt injection.
The workflow explicitly requires manual review of notebook content and gives no untrusted-content boundary. No embedded attack is present in the skill files themselves.

Remediation

Suggested fixes recorded by this audit. Applying them is the maintainer’s responsibility.

  1. FIX-001
    High
    Notebook and project paths are interpolated into a shell command.
    Invoke detect-secrets with a direct argument list. Remove sh, echo, tr, and xargs, and pass every path as a separate argument.
  2. FIX-002
    Medium
    The validator resolves and executes unpinned nbconvert and detect-secrets packages.
    Pin reviewed package versions and hashes. Disclose network access and require confirmation before resolving missing dependencies.
  3. FIX-003
    Medium
    The workflow asks the agent to read notebook-derived Markdown without treating notebook content as untrusted.
    State that notebook text is audit data. Instruct the agent to ignore embedded commands, authority claims, and requests unrelated to the audit.

Expert evidence

Immutable subject identity, scanner metadata, dismissed matches, and source-level evidence.

Artifact subject

Marketplace commit
d71c7417a35d5c2624161bd2fe8de8a41a362128
Content hash
63b4c68ffe7ea6c95b13ab008d58224651a5c63d3559ae4e58ec317f72a16258
Tree hash
0e576b70a1c58a4d126c2cc04e6376f7c5f28a7803bfa5c2603cba790bd29d5a
Skill path
skills/anthropics/cookbook-audit
Audit payload hash
098b38cf25cbf7a5dc9fe8b8a5c1db13

Analysis metadata

Audit model: codex

Analysis state: Complete

Scope is limited to the recorded files, lines, methods, and evidence. No runtime or sandbox execution is claimed.

Verify and export

JSON report CITATION.cff SARIF 2.1.0 Signed installation manifest Source

The manifest and lockfile bind install artifacts to cryptographic hashes. This integrity claim is separate from the security assessment.

Audit attestation: not_attestable