Skills actionbook

📦

actionbook

Name: actionbook
Author: actionbook

v1.5.0 Content revision r1 High Risk ⚙️ External commands🌐 Network access📁 Filesystem access🔑 Env variables

Automate Browser Workflows with Actionbook

Complex browser tasks require current selectors and reliable session control. This skill uses Actionbook snapshots and explicit sessions to automate websites.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Stop for confirmation before installing.

Review the plan and obtain explicit user consent before changing files.

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "actionbook" from https://skillstore.io/skills/actionbook-actionbook.md and its manifest at https://skillstore.io/api/skills/actionbook-actionbook/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/actionbook-actionbook.md Signed manifest GET /api/skills/actionbook-actionbook/manifest Signed lockfile GET /api/skills/actionbook-actionbook/lockfile

Test it

Using "actionbook". Inspect the product page and summarize availability.

Expected outcome:

The page loaded successfully. Three variants are available, one is unavailable, and no cart or checkout action was performed.

Using "actionbook". Fill the registration form but do not submit it.

Expected outcome:

The requested fields are complete. The form remains open at the final review step and awaits confirmation before submission.

Using "actionbook". Compare search results in three tabs.

Expected outcome:

Tab one returned 18 matching items.
Tab two returned 12 matching items.
Tab three requires manual login before results are visible.

Security Audit

High Risk

v2 • 8/15/2026 Open versioned report

Most alerts are false positives caused by Markdown, example domains, and benign browser terminology. Confirmed risks include browser evaluation, credential access, and sensitive traffic exports. The skill also lacks confirmation boundaries for consequential actions and browser credential access.

Files scanned

850

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (5)

High

references/authentication.md:145

Environment file access

1. **Use environment variables for credentials** — source from a `.env` file, not inline

The guidance explicitly recommends sourcing credentials from a .env file. This exposes secrets to the shell environment and treats executable shell input as configuration.

High

references/authentication.md:148

Environment file access

source .env # contains APP_USERNAME, APP_PASSWORD

The command sources .env directly, which executes the file as shell code and loads username and password values into the process environment.

High

references/command-reference.md:296-318

Sensitive Browser Traffic Export

Request dumps include response bodies, while HAR files include headers. Temporary and default paths lack documented permission controls or secret redaction.

The reference explicitly describes exported response bodies and HAR headers at local paths. These commonly contain credentials or private application data.

High

references/command-reference.md:343-371

Direct Access to Browser Credentials and Storage

Commands can list cookies and read local or session storage without requiring user confirmation, domain restrictions, or output redaction.

The command reference directly exposes cookie listing and storage retrieval, including examples named session_id and auth_token. No access guard is documented.

High

SKILL.md:21-28

Missing Confirmation for Consequential Website Actions

The skill activates for messages, bookings, posts, and form submissions but does not require confirmation before external communications or transactions.

The activation list explicitly includes messaging, booking, posting, and form submission. The documented workflow has no confirmation gate before consequential actions.

Capability review items (15)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

High

references/command-reference.md:312

Hidden file in home directory

actionbook browser network har stop --session s1 --tab t1 # Stop and export

The command exports a HAR file under the hidden ~/.actionbook directory. HAR headers can contain sensitive authentication and request metadata.

High

references/command-reference.md:316

Hidden file in home directory

Recording is per-tab: multiple tabs (or sessions) can record independently at the same time. `har st

The documentation confirms that HAR output is written to ~/.actionbook/har when no path is supplied. The trace includes request and response headers.

High

references/command-reference.md:66

Generic API/secret keys

Supported cloud providers: `driver` (`DRIVER_API_KEY`), `hyperbrowser` (`HYPERBROWSER_API_KEY`), `br

The cloud provider modes read named API keys from the shell environment. This is intentional secret access by the external Actionbook process.

High

SKILL.md:12

Generic API/secret keys

- name: ACTIONBOOK_API_KEY

The skill declares ACTIONBOOK_API_KEY as an environment variable consumed for authenticated service access. This is intentional handling of a secret.

High

SKILL.md:102

Generic API/secret keys

Use `-p` / `--provider` with `browser start` to run sessions on a remote browser instead of launchin

The text states that each remote browser provider reads its API key from the shell environment. This is explicit secret access by the workflow.

Medium

references/command-reference.md:312

Hidden file access

actionbook browser network har stop --session s1 --tab t1 # Stop and export

This command writes captured browser traffic into a hidden home directory. The output can contain sensitive request headers and browsing metadata.

Medium

references/command-reference.md:316

Hidden file access

Recording is per-tab: multiple tabs (or sessions) can record independently at the same time. `har st

The stated default writes HAR traces beneath ~/.actionbook/har. This is real hidden-directory access involving potentially sensitive browser traffic.

Medium

references/command-reference.md:296

Temp directory access

actionbook browser network requests --dump --out /tmp/dump --session s1 --tab t1 # Export matching

The command exports captured requests and response bodies to /tmp/dump. Shared temporary storage may expose authenticated browser data to other local users.

Medium

references/command-reference.md:297

Temp directory access

actionbook browser network requests --dump --out /tmp/dump --filter /api/ --session s1 --tab t1 # E

The filtered dump still writes request records and response bodies under /tmp. The documentation does not require restrictive permissions for this sensitive output.

Medium

references/command-reference.md:313

Temp directory access

actionbook browser network har stop --session s1 --tab t1 --out /tmp/trace.har # Stop and export

The command writes a HAR trace to /tmp/trace.har. HAR request and response headers can contain tokens or other sensitive metadata.

Medium

SKILL.md:128-129

Ruby/shell backtick execution

`browser eval` accepts the expression from three mutually-exclusive sources:

Although this is not Ruby backtick execution, it introduces browser eval, which executes arbitrary JavaScript in the active page context.

Medium

SKILL.md:129-130

Ruby/shell backtick execution

- **Positional**: `actionbook browser eval "expr" ...`

The documented positional form executes an arbitrary JavaScript expression in the active page. That code can access page data and authenticated requests.

Medium

SKILL.md:130

Ruby/shell backtick execution

- **`--file`**: `actionbook browser eval --file script.js ...`

The --file form reads and executes a local JavaScript file in the browser context. This is real dynamic code execution despite the inaccurate pattern label.

Medium

SKILL.md:131-135

Ruby/shell backtick execution

- **Stdin**: `echo 'expr' | actionbook browser eval - ...`

The stdin form executes piped JavaScript in the active page. Untrusted upstream input could therefore become browser-context code execution.

Low

references/command-reference.md:144

Fetch API call

actionbook browser eval "await fetch('/api/data').then(r => r.json())" --no-isolate --session s1 --t

The example executes fetch inside the active page. It is an intentional same-origin network request, but it can transmit authenticated browser data.

Risk Factors

⚙️ External commands (50)

🌐 Network access (25)

📁 Filesystem access (9)

references/authentication.md:156 references/authentication.md:156 references/command-reference.md:312 references/command-reference.md:316 references/command-reference.md:312 references/command-reference.md:316 references/command-reference.md:296 references/command-reference.md:297 references/command-reference.md:313

🔑 Env variables (4)

references/command-reference.md:66 SKILL.md:12 SKILL.md:102 SKILL.md:105

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/actionbook-actionbook/audits/2?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/actionbook-actionbook/security.svg)](https://skillstore.io/skills/actionbook-actionbook?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/actionbook-actionbook?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/actionbook-actionbook/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/actionbook-actionbook.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

actionbook. (2026). actionbook security audit report (audit version 2) [Author version 1.5.0]. Skillstore. https://skillstore.io/skills/actionbook-actionbook/audits/2

BibTeX citation

@techreport{actionbook-actionbook-actionbook-2026,
  author = {actionbook},
  title = {actionbook security audit report (audit version 2)},
  institution = {Skillstore},
  year = {2026},
  number = {2},
  url = {https://skillstore.io/skills/actionbook-actionbook/audits/2},
  note = {Author version 1.5.0}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "actionbook security audit report (audit version 2)"
version: "1.5.0"
type: report
authors:
  - name: "actionbook"
date-released: "2026-08-15"
url: "https://skillstore.io/skills/actionbook-actionbook/audits/2"
identifiers:
  - type: other
    value: "skillstore:actionbook-actionbook:audit:2"
    description: "Skillstore immutable audit report identifier"

Skillstore Score

Why this score Evidence Confidence: Medium

Architecture

100

Maintainability

Content

Community

Spec Compliance

What You Can Build

Automate Repetitive Website Tasks

Navigate forms, enter approved data, and complete repeatable browser steps with current element references.

Build Browser-Based Tests

Run navigation, interaction, waiting, screenshot, log, and network checks across controlled test environments.

Collect Structured Web Results

Open multiple pages, snapshot current content, and extract requested text from approved websites.

Try These Prompts

Inspect a Web Page

Open [URL], create a fresh session, snapshot the page, and summarize the visible sections. Do not submit any forms.

Complete a Form Safely

Open [URL], snapshot the form, fill fields from [DATA], and pause for my confirmation before submitting.

Compare Results Across Tabs

Open [URLS] in separate tabs, apply [CRITERIA], and compare visible results. Report blocked pages and request login when required.

Run a Browser Test Workflow

Test [WORKFLOW] on [TEST_URL]. Capture failures, console errors, and screenshots. Redact secrets and avoid production-changing actions.

Best Practices

Snapshot each page before interacting, and refresh references after navigation or major page changes.
Use separate sessions or profiles for unrelated sites and close sessions after completing the task.
Confirm consequential actions and redact credentials, cookies, storage values, and network data from outputs.

Avoid

Do not reuse remembered selectors when a current snapshot is available.
Do not automate login challenges, MFA, CAPTCHA, purchases, posts, or submissions without user participation and confirmation.
Do not evaluate page-derived scripts or export sensitive traffic into shared temporary directories.