Back to skill

Security audit

Info Visualize

Security checks for vulnerabilities and agentic risk

Overview

The skill is a coherent chart/report generator, but its HTML/SVG output can include unescaped input data, creating a script-injection risk when rendering untrusted data.

Install only if you will render trusted data or review/sanitize JSON before generating HTML charts. Be careful publishing generated HTML/SVG from untrusted inputs, prefer PNG output for untrusted data, pin Pillow in a virtual environment, and use --archive only for reports you intentionally want copied to another directory.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (3)

T01 · Skill Instruction Hijacking

Note
Location
scripts/svg_bar_chart.py:242
Finding

Unavoidable Third-Party Branding in Generated Artifacts

Content
View full analysis
{footer} Powered by WorkBuddy ``` From `scripts/png_longform.py:385-388`: ```python # WorkBuddy watermark wm = "Powered by WorkBuddy" wm_w = draw_obj.textbbox((0, 0), wm, font=F_SMALL)[2] draw_obj.text((self.W - wm_w - PAD, y0 + 22), wm, font=F_SMALL, fill=(38, 58, 82)) ``` ### Technical Analysis Both renderers unconditionally add the fixed text `Powered by WorkBuddy` to generated artifacts. The Skill presents itself as a general visualization utility, but it provides no documented or implemented option to disable or replace this attribution. This behavior modifies user-requested output with unrelated fixed branding. It does not grant code execution or system privileges, but it can affect the integrity and provenance of reports intended for publication. ### Attack Path 1. A user loads the Skill and requests an SVG/HTML chart or PNG report. 2. The relevant renderer processes the supplied report data. 3. The renderer unconditionally inserts the WorkBuddy watermark. 4. The user publishes or distributes the artifact without realizing that third-party branding was embedded. ### Impact Assessment No operating-system privileges, credentials, or data-access capabilities are obtained. The impact is limited t ...[truncated 157 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/svg_bar_chart.py:130
Finding

Stored HTML and SVG Injection Through Unescaped Chart Data

Content
View full analysis
{kpi['value']} {kpi['label']}""" ``` Item labels, tooltips, and tags are inserted without escaping at `scripts/svg_bar_chart.py:160-170`: ```python rows_svg.append(f""" {i+1:02d}. {label} {tooltip} {display_val} {tag}""") ``` The chart title, footer, and HTML document title are also interpolated directly at `scripts/svg_bar_chart.py:217-218`, `242-253` ...[truncated 3079 chars]
Remediation
View remediation
``` - Add regression tests using payloads containing `

T08 · Insecure Dependencies

Note
Location
SKILL.md:22
Finding

Unpinned Pillow Dependency in Installation Guidance

Content
View full analysis
Remediation
View remediation
``` - Generate and verify cryptographic hashes, for example through a hash-locked requirements file: ```bash pip install --require-hashes -r requirements.txt ``` - Commit the dependency manifest or lock file to the project. - Use a trusted package index and avoid unreviewed mirrors. - Regularly update the pinned version after vulnerability review and compatibility testing. - Update `SKILL.md` and the import error message to direct users to the project dependency file rather than an unconstrained installation command. - Run the renderer in an isolated virtual environment with only the permissions required to read its input and write its output. ]]>
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (7)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
88% confidence
Finding

The skill documents capabilities that read input files and write output files, but it does not declare any explicit tool scope or permissions boundaries. This can lead to overbroad execution in an agent environment, where the skill may access or modify files without clear policy enforcement or user awareness.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger phrases are broad and match many ordinary requests such as 'make chart' or 'visualize data', which can cause the skill to activate in contexts the user did not intend. Overbroad activation is risky because the skill performs file generation and may write outputs automatically once selected by the agent.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The documentation normalizes writing report files and archiving copies without warning the user about filesystem side effects. In an autonomous agent setting, undisclosed persistence can leak sensitive data into local storage or retained directories and make cleanup difficult.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
77% confidence
Finding

The natural-language description and examples are presented as a Chinese-only workflow, and the skill positions itself as a general visualization skill rather than a clearly region-specific tool. There is no statement that output language is user-selectable or that Chinese is merely the default.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The skill includes an archive-copy behavior to a user-specific memory/news directory that goes beyond simple visualization output. Even if the behavior is convenience-oriented, extra persistence outside the immediate output path increases data exposure and can surprise users by storing generated content in secondary locations.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The manifest describes rendering structured data into charts and long-form PNG reports. The optional --archive flow creates directories and copies the generated PNG into an arbitrary archive location, which is a file-management capability not necessary to perform visualization itself.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The generated document hard-codes lang="zh-CN", which imposes a specific language/locale choice in output regardless of user preference. This is a natural-language policy concern because the skill does not offer any language or locale selection or document a justified region-specific constraint.

Content

No source excerpt is available for this finding.